// transcript — 3938 segments
0:00 Introduction
0:02 In a world being transformed by AI, one skill stands above all others. AI
0:07 automation. Master this and you won't just survive the AI revolution, you'll
0:11 thrive in it. I'm living proof of this. Just 2 years ago, I taught myself how to
0:15 build no code AI automations without any prior experience. And since then, I've
0:19 built multiple AI businesses, generated millions of dollars in revenue, and
0:22 grown this channel to over 500,000 subscribers, and built AI systems for
0:25 some of the biggest brands in the world. It's pretty safe to say that learning
0:29 how to build AI automations has completely changed my life. So, in this
0:32 full course, I'll teach you everything that I've learned about building AI
0:35 automations and making money with them, even if you don't know how to code. And
0:38 as they say, AI will not replace you, but the person using AI will. So, my
0:42 hope is that with this video, you too can learn this incredibly powerful skill
0:46 to build the life of your dreams before it's too late. And as you can tell by
0:49 the length of this video, I'm not going to be holding anything back. So, I've
0:52 split it into three different chapters. Firstly, we'll build your foundational
0:56 understanding of AI automation, covering what it actually is, the different types
1:00 of AI automations, how they work under the hood, and the key concepts you need
1:04 to know before we start building. There's no technical background required
1:07 to understand any of what I'm going to teach you there. Secondly, we'll dive
1:10 deep into building out actual AI automations, taking you over my shoulder
1:13 every step of the way as we build some of the most in- demand AI automation use
1:17 cases in the market today. This includes building things like cuttingedge voice
1:21 agents, too. And in the third and final chapter, I'll be giving you my proven
1:24 blueprint for monetizing your AI automation skills while this technology
1:28 explodes. I'll share the exact strategies that I've used to generate
1:31 millions of dollars with the skill set. So, if you're new to the channel and
1:33 don't know who I am, let me quickly share why I am qualified to teach you
1:36 about AI automations in the first place. So, my name is Liam Mley and just 2
1:41 years ago, I started learning AI with no prior experience in the field. Teaching
1:44 myself how to build AI automations and chatbots through my own self-study,
1:47 which I documented here on this YouTube channel from day one. This led me to
1:51 starting Morningside AI, my AI automation agency, where we build AI
1:55 systems and agents for businesses from basic customer support systems when we
1:59 started to now full AI SAS platforms for some of the biggest brands in the world.
2:01 And I also have my own AI SAS called Agentive, which has over 70,000 users on
2:05 it. At Morningside AI, we've worked with publicly traded companies and even an
2:09 MBA team recently. And I also run the world's largest AI automation and
2:12 business community with over 180,000 members on school. So through this
2:15 community and my YouTube channel, I've taught hundreds of thousands of people
2:19 from all backgrounds how to build and make money from AI automation. And
2:22 everything I'm about to teach you today is exactly what helped me to achieve all
2:26 of this. So let's dive in. So there's a lot to cover here. I don't want you to
2:29 give up halfway. So let's quickly get clear on why learning AI automation is
2:33 one of the most valuable skills anyone can have over the coming decade. Whether
2:36 you're a student, an employee, or an entrepreneur. Here's some quick truths
2:40 about AI and jobs. McKenzie predicts that AI and automation can replace up to
2:45 50% of current work activities by 2030. And the World Economic Forum states that
2:49 41% of companies plan to reduce staff due to AI. Now, this is a lot of doom
2:52 and gloom and many are naturally worried about their career in the future when
2:55 they hear the stuff, but it's not actually all bad if you know where to
2:58 look. So, on the flip side of this same data, these same reports reveal an
3:02 enormous opportunity for those willing to seize it. The board economic forum's
3:06 future of job report states that 50% of employees plan to reorient their
3:09 business in response to artificial intelligence and 66% of employees plan
3:15 to hire talent with specific AI skills such as AI workflow automation. So on
3:18 one hand we have the expectation of massive layoffs and automation of work
3:22 over the next 5 to 10 years. But on the other we have the majority of employers
3:26 searching for people who have AI skills or really just some form of basic AI
3:29 literacy. Why is this? Well, it's because AI literate individuals who can
3:33 identify opportunities for automation and automate them themselves can have 5
3:36 to 10x the output of someone who doesn't know this and can't automate their own
3:40 work. And I promise you that brushing up on your AI and actually becoming AI
3:43 literate so that you can be on the winning side of this next 5 to 10 years
3:47 is so much easier than you think. I mean, it's it's literally as easy as
3:51 watching this entire video in order to build your AI skills base. If you don't
3:54 believe me when I say that a little bit of self-study like this video goes a
3:57 long way, here is an excellent clip from the All-In podcast from one of the most
4:01 respected investors and technologists in the world, Naval Ravakant, alongside a
4:05 whole bunch of other biners. Again, I would say the easiest way to see that AI
4:08 is not taking jobs or creating opportunities is go brush up on your AI,
4:14 learn a little bit, watch a few videos, use the AI, tinker with it, and then go
4:18 reapply for that job that rejected you and watch how they pull you in. This
4:21 video is exactly what Naval is talking about. This is why I create these. So
4:24 whether you're a student wanting to stand out in a competitive job market or
4:28 an employee aiming to become irreplaceable at work or an entrepreneur
4:31 like me looking to scale your business with cuttingedge tools and automations,
4:35 I have made this video for you. Now close out of all your other tabs, get a
4:38 notebook and a pen and a beverage of your choice and make a commitment right
4:41 now to yourself to finish this training and to ensure that you're going to be
4:45 empowered by AI and not replaced by it. That is all I want out of this video for
4:47 you all. So if you've done all that, let's get stuck into it.
4:53 All right. So, step one in building AI automations is knowing what an
4:56 automation actually is. And the term gets thrown around a lot these days.
4:59 First thing we need to realize is that the AI part of the term is is relatively
5:03 new. Automation itself has been around for a long time. So, let's start there
5:07 to make this super easy to grasp. In simple terms, an automation is a system
5:10 that does a task for you without you having to lift a finger. It's kind of
5:13 like setting up a little robot to do boring, repetitive stuff automatically,
5:16 so you don't have to waste your time on it. These are what we'll call old school
5:20 automation. The kind that existed way before chat GBT came along. They were
5:23 often built on platforms like Zappia and lots of small to medium-sized businesses
5:27 used them for the past 5 10 years. And they did basic things like automatically
5:31 saving info, for example, when someone filled out a form on a website. You
5:34 could make an automation that would take their name and their email and then just
5:37 pop it into a spreadsheet. So just a little automation automating that boring
5:40 stuff. Or for example, when an email came in, it would send a quick alert to
5:44 a chat app like Slack. So it's kind of like having a little helper who's
5:46 following a super simple checklist. If this happens, then do that. No thinking,
5:50 just just doing, right? And the benefit of this is huge. It freed humans from
5:54 doing super basic and boring work. And for business owners, it meant not having
5:57 to pay more people just to handle these tiny and annoying tasks. It's kind of
6:01 like having a tireless assistant who never complains about doing the same
6:04 boring thing over and over and over again. So for decades, all was well in
6:08 the automation space. And these old school automation saved time and money,
6:11 and everyone was super happy. That was until the release of Chat GBT in late
6:16 2022. It blew the entire field wide open, turning automation from some niche
6:20 trick used by some savvy companies into the biggest thing since the internet.
6:24 Generative AI models like chat GBT added to automation was like putting a V12
6:27 onto a bicycle for this automation space. They made it possible to do more
6:31 than just simple tasks. This was because these powerful AI models could handle
6:34 much trickier stuff that used to require a human brain. Instead of just updating
6:38 a spreadsheet row automatically, these automation platforms can now use the
6:42 power of Chat GBT to do things that only people could have done before. So
6:45 platforms like make.com, which you'll be learning more about later in this video,
6:48 enable us to automate things like write a whole post for LinkedIn sounding just
6:52 like you. Pulling out names, places, and phone numbers from giant documents in
6:55 seconds. Reading and figuring out if an incoming email is someone asking for a
6:59 refund or just wondering where their order is. Shrinking huge piles of info
7:03 into short and easy to read reports, spotting things in the picture like
7:06 identifying a product in a photo, or even creating brand new images and
7:10 videos from just a few words. So what chat GPT and the explosion of other
7:14 amazing generative AI tools gave us was basically human intelligence on demand.
7:17 These AI models are kind of like having a super smart friend who can do almost
7:21 anything that you ask them as long as you tell them clearly what you want. And
7:24 using automation platforms, we can easily set up these super smart friends
7:28 into our systems and use these kinds of models in thousands of different ways.
7:32 All you have to do is pick the right AI tool for the job, give it a clear
7:35 instruction, call a prompt, and watch the magic happen before our eyes. And
7:38 that is how the AI automation industry was born. So, with that little history
7:42 lesson out of the way, let's get back to our original question of what is an AI
7:46 automation. Well, it turns out that this field is so new that there isn't even an
7:49 official definition for what an AI automation is. So, here's mine. Just
7:53 keeping it nice and simple. An AI automation is a system that uses AI to
7:57 automatically do complex tasks that would normally require a human. So, the
8:00 big difference between these old school automations that we've just talked about
8:04 and today's AI automation is the kinds of tasks that they can handle. Thanks to
8:08 these recent advances in AI technology, we've gone from just moving data around
8:11 and putting stuff in spreadsheets to being able to solve problems that need
8:14 thinking and creativity and decision-making. It's kind of like
8:17 upgrading from a basic toy robot that only moves forward to a high-tech robot
8:21 that can solve puzzles and move around the world. So, when you learn AI
8:24 automation, you are basically learning how to build digital workers that can do
8:27 very powerful things for you without ever having to lift a finger. This is
8:31 why so many people are racing to pick up the skill right now before it's too late
8:34 to stay ahead. It's like the ultimate cheat code because you can build them to
8:38 do exactly what you need, tailored to any kind of job or workflow. For
8:41 example, for students, it's kind of like having a helper that can organize your
8:44 study notes automatically. Or for employees, your AI automations can be
8:47 like a buddy that handles the boring paperwork so you can shine on bigger
8:50 projects. And for entrepreneurs, it's like having a team member who runs your
8:53 business tasks while you sleep. And the great thing is that they all cost way
8:56 less than hiring extra people. And they don't need breaks or vacation days, and
8:59 they don't grumble about doing the boring stuff. So, I'm sure you can see
9:03 why businesses and students and anyone looking to save time and money are so
9:07 excited about this technology and why anyone who knows how to build these
9:11 systems is instantly 10x more valuable. Just imagine automating something simple
9:15 in your life or work like sorting emails or scheduling tasks and how much time
9:19 you'd be able to save to focus on what really matters in your life.
9:24 Now, before we dive deeper, it's important to understand that AI
9:27 automation is a super broad term these days, covering wide ranges of different
9:31 systems and applications that can be built with AI. This is largely due to
9:34 the rapid advances in areas like AI agents and AI tools. So, over the past 2
9:39 years, as I built my own AI agency, Morningside AI, and helped thousands
9:42 through my communities to do the same, I've had to create a clear system for
9:46 making sense of the chaos that is the AI automation landscape. And I really want
9:49 to share that with you today because it's been super helpful for me and many
9:52 of my students to be able to figure this space out and at least have some mental
9:55 buckets that you can put things in. Here's the three different categories
9:57 that you need to keep in mind. And please stick with me. This will all make
10:00 sense in a second. So firstly, we have conversational AI. These are systems
10:03 that chat with people handing back and forth conversations. It's kind of like
10:07 having a friendly robot that talks to customers for you. These kinds of chat
10:09 bots can be found on things like websites and answer questions or they
10:13 can be voice agents that pick up phone calls. These used to need real people to
10:16 talk, but now AI can automate these kinds of conversations. The second
10:19 category is AI tools, and these are systems that use AI to do a specific job
10:23 when a person asks them to, and it's mostly to help workers get more done.
10:27 For example, I can make a custom AI tool that takes a link to a cool blog post
10:31 that I found, grabs the info from the web page, does extra searches on the
10:34 topic, and then uses something like chatbt to write a new beta version for
10:39 my own blog. And third and final is AI workflow automations. These are systems
10:43 that do a whole series of tasks by themselves, starting when something
10:46 happens, like a trigger or on a set schedule like once a day. They use AI to
10:50 make decisions that used to need a human brain. It's kind of like having a smart
10:53 robot manager that runs the whole process for you. For example, an
10:56 automation can call customers of an online store 14 days after they buy
11:01 something using an AI voice agent to ask for feedback and review all without you
11:04 having to do a thing. So, you set it up on that trigger of 14 days after
11:08 purchase, then execute this workflow. Now, those three categories might be a
11:10 little bit confusing if you are completely new to the space, but don't
11:13 worry. In the building section of this video, we're going to be creating three
11:16 automations which integrate each of these different types, and you get to
11:19 see them in action, which will make it super clear. So, AI automation is
11:22 essentially an umbrella term under which all of the exciting stuff in the AI
11:25 space is happening right now. However, when people talk about AI automations,
11:28 they are typically referring to the last type that I mentioned. So, AI workflow
11:31 automation, and these are what we're going to be building later in the video.
11:35 And an automation refers to one chain of steps that uses AI in various ways to do
11:39 certain tasks. And if I'm honest, this last category of AI workflow automation
11:42 is really the most powerful because it can incorporate all elements of AI
11:46 automation like agents, conversational AI, and tools in order to build
11:50 end-to-end processes that are much more valuable than if they were just alone.
11:52 So, what I'm saying is that what you're about to learn is the key skill that
11:55 lets you do pretty much anything in the AI space these days. It is the
11:59 foundation for building systems that save time, make money, or just make life
12:03 easier, whether you're at school or work or running your own show. And after
12:05 years in the game, I can tell you that it's been one of the most valuable
12:10 Anatomy of an Automation
12:10 skills that I have ever picked up. So, now that you understand what AI
12:13 automations are, let's take a little peek under the hood and see how they
12:16 actually work. So, don't worry if this sounds tricky. I've been breaking down
12:19 this kind of complex AI stuff for years now. So, I'm going to make this super
12:21 easy to understand for you. So, you can think of an AI automation like a facto's
12:25 assembly line, right? There's different stations and they're all working
12:28 together to build something awesome from start to finish. It's kind of like
12:31 having a a team of little robots, each with a special job passing the project
12:35 along until it's done. So, let's go through the five key parts to make this
12:38 magic happen. Firstly, we have the trigger. This is the very first step of
12:41 an automation. You can think of it as the facto's start button or the whistle
12:45 that says, "Let's go." It's what kicks everything into gear. It could be
12:48 something like a new email popping into your inbox, a form being filled out on a
12:53 website, or even a specific time of day. This is more of a a schedule. Secondly,
12:57 we have a filter. So, not everything that starts in the automation should
13:00 keep going through it. So, a filter essentially checks if what came in is
13:03 the right stuff to work on. It's like how a factory worker does some kind of
13:06 quality control and checks that if the materials that they've received are good
13:10 enough to use in the final product. If they are not then they get tossed out
13:12 and if they are then they move forward to the next part of the sequence. You
13:15 can think of it kind of like a bounce at a club where only the important stuff
13:18 and the good things that you want inside the club or in your automation are
13:21 allowed through. Thirdly, we have actions. So this is where the real work
13:25 gets done. Actions are the steps your automation takes like the different
13:28 stations in a factory where each one does a specific job. So for example,
13:31 your automation might send an email, update a list or create some kind of
13:34 report. Often there are going to be several actions one after another just
13:37 like a product moving down the line getting built bit by bit in order to
13:41 achieve one of these outcomes. Next we have the intelligence layer. So this is
13:44 where the AI magic shines. This part is like having a super smart robot on the
13:48 assembly line that can think, analyze, and make decisions on the spot. And you
13:51 can tell it how to think using prompting. The AI inside your automation
13:55 can look at each task, figure out what's needed and adapt based on the context
13:58 you provided, like deciding how urgent something is, writing a custom message,
14:02 or pulling out key info from a big mess of data. These intelligent steps can go
14:06 way beyond just following some kind of preset rules that we saw with old school
14:09 automations. Next, we have the format. So, just like items in production line
14:13 may need some sanding or trimming before another piece can be added on top, the
14:16 data in our automations often need some kind of adjustments along the way. And
14:19 this is where we use a formatter to prepare things for the next step. And
14:22 finally, we have the output. This is your finished product, just how Carac
14:25 Factory packages up the final item and ships it out. Your automation delivers
14:28 the completed work at the end of the sequence. This could be a message sent
14:32 to your team, an updated file in your system, or a finished document that's
14:35 ready to go. At the end of the line, it's where everything comes together.
14:38 It's like after making pizza and putting all the toppings on and prepping the
14:40 dough, you're finally getting out of the oven. It's hot. It's ready to eat and
14:43 ready to go. So, let me show you how all of these parts work together with a real
14:46 example that anyone in a job can relate to. Say you're an employee who wants an
14:50 automation to handle incoming customer emails while at work. So the trigger is
14:53 going to be when a new email lands in your company's support inbox. The filter
14:56 is going to check if it's something important like if the email mentions
15:00 urgent or problem then it will pass it along to the next step. If not, it will
15:03 end the sequence there. Thirdly, we have the intelligence layer which is going to
15:06 use AI to read the email, figure out what it's about and then helps to draft
15:10 a helpful response. And in this case, the intelligence layer also acts as a
15:13 formatter which is essentially packaging the response and making sure it's in the
15:16 format we want. Then the actions in the automation can actually send that reply
15:20 and alert the boss on Slack if needed. The output logs everything neatly in
15:24 your system and marks it as handled. So that's just a simple example, but these
15:27 same building blocks are used to automate much bigger things too, which
15:31 is why AI automations are such a game changer these days. They combine these
15:34 parts to create truly smart systems that can save time and boost efficiency,
15:38 whether you are you're juggling tasks at work or crushing it in school or running
15:40 your own business. So, now that you've got a handle on what AI automations are
15:44 and how they're built, let's take a quick look at the tools that make all of
15:48 Exploring the Ecosystem
15:48 this possible. Now, don't worry if this sounds a little bit too techy. Soon, these tools are
15:53 going to literally be second nature for you, and it's going to feel so easy to
15:56 do. Creating automation starts with picking a main automation platform.
15:59 These platforms are typically called workflow builders, and they are the
16:03 command center of your automation factory. They give you a blank canvas to
16:06 design your automation on using the building blocks we talked about. Some of
16:09 which are going to be powered by AI like chatb. Popular workflow builders include
16:13 make.com which we're going to be using later in this video in the tutorial
16:16 section. We have Zapia which is great for quick setups and we have NAM which
16:19 is perfect if you want a bit more control. They are the brain of your
16:22 operation and they're really just controlling how everything fits
16:25 together. So workflow builders don't work alone. They are essentially a place
16:28 to hook up all sorts of other tools to get the job done. So these are the
16:30 categories of tools which you can connect to your workflow builders.
16:33 Firstly, we have databases and spreadsheets. So for storing information
16:36 and data, you can use things like Air Table or Google Sheets. You can think of
16:39 them as basically the filing cabinets where you keep all your data neat and
16:42 tidy so that you can save new things to the database or you can pull it into
16:45 your automations as needed. Secondly, we have communication tools for sending
16:49 messages. Things like Slack or Google. They're like kind of walkie-talkies that
16:51 can pass information around automatically for you. Then we have AI
16:55 models. And these can add that smart human level thinking like open AI's chat
16:58 GPT which you can think of kind of like a genius buddy who solves problems for
17:02 you when you give him the instructions that he needs. Then we have scheduling
17:05 tools which can handle time and meetings things like calendarly or Google
17:07 calendar and these are essentially like your personal planner to keep things
17:10 running on schedule. Then we have forms and intake too. So this can collect
17:13 information from people where you have things like type form or tally. These
17:17 are essentially input forms and triggers for your automations when someone fills
17:19 them out. So as your automation goes through it's like calling up all your
17:22 friends on a group project. Each one is bringing their own special knowledge or
17:26 capabilities to the table in order to help you to get to the end of that
17:28 sequence and finish off the automation to create the final product. By building
17:31 out a workflow, you are basically the boss at the factory. Once you know what
17:35 each tool can do, you can mix and match them to work together smoothly along
17:38 this assembly line. It's kind of like building with Lego blocks. You can just
17:42 like snap the right pieces together to make something awesome. You get to
17:45 decide which tools you connect in what order and how AI can make the whole
17:49 system smarter as well. Which AI models do you use? Gemini, do you use OpenAI?
17:52 to use complexity for searching. There's so many different decisions you have to
17:55 make. Uh but it's really cool to be able to flex your kind of creativity in order
17:58 to solve these problems. It's really like a a new age form of problem
18:01 solving, which is why I love it so much. I say, I know I need to take this and
18:05 get to this. How can I use AI to do that? And really forces you to explore
18:08 what's out there in terms of AI tools these days. So, what kinds of workflows
18:11 can we build? So for students, for example, you could build a study
18:14 material organizer that automatically summarizes lecture recordings for you,
18:18 creates flashcards from notes, and schedules review sessions based on your
18:21 exam dates. An employee could improve their workflow with a meeting assistant
18:24 that records and transcribes their meetings, and then generates action item
18:28 summaries off that and updates project management tools. Entrepreneurs could
18:31 design a lead qualification system, which automatically qualifies leads,
18:35 calls them with an AI voice agent, and then sends them a custom proposal to
18:38 automate that whole process. And that is exactly what you'll learn how to build
18:41 by following along with me in this next section. So, I'm going to show you how
18:43 to build out a lead qualification automation step by step, starting simple
18:48 and getting more complex as we go. We're going to be focusing on this
18:50 business-based workflow because these kinds of business systems offer the best
18:54 opportunity for monetizing your new skills, which I'll go into depth on at
18:57 the very end. So, stick with me and you're going to be learning how you can
18:59 start to make money immediately with these kinds of skills. So, before we get into that, let's do a
19:01 Foundations Recap
19:05 quick summary of this section. So firstly, an AI automation is a system
19:09 that uses AI to automatically perform complex tasks that used to require
19:14 humans. Secondly, we build them inside of workflow builders that incorporate
19:17 integrations with tons and tons of other tools. And every automation has six key
19:21 components. Firstly, a trigger, what starts the workflow, a filter,
19:25 conditions that need to be met, the intelligence layer, which processes info
19:29 and makes decisions with AI, the actions, which are the tasks that
19:32 actually get performed, formatterers, which clean things up, and the output.
19:36 the final result or deliverable. So, if you're feeling unclear about anything
19:39 we've covered so far, feel free to go back and rewatch some of those sections
19:42 and be ready to join us in the next step when we begin building. So, it's really,
19:45 really important that you do understand everything that we've gone over there
19:48 because you're not going to have a solid foundation to build your technical
19:51 skills on top of, which we're doing in the next section. So, please, please,
19:54 please, I've taken a lot of time to put all of this information very, very
19:57 gradually together. So, you must understand everything that I've gone
20:00 over just here before we go into the next phase. If that's all good, then
20:02 let's take a look at what we're going to Okay, so now that we have the
20:08 foundational knowledge built that you need, we can now get into the second
20:11 chapter of this video where we're going to be building three AI automations from
20:14 scratch. We're going to be starting off with something super beginner friendly
20:17 to get you started and then working our way up to a much more advanced and
20:20 valuable one by the end. Now, very important note is that each of these
20:23 automations build on each other. So, you have to be able to do the first one in
20:26 order to be able to make the second and so on. So, you you cannot skip ahead.
20:30 I've planned this out very carefully to gradually layer on your skills. So, it's
20:33 all very intentional. So, please just trust the process. So, over the next
20:36 chapter, you'll learn almost all of the key skills you need to start building
20:39 your own AI automations from scratch and be able to tap into this enormous
20:42 opportunity that is AI automation. The system that we're going to be building
20:45 gradually over the next three sections is an AI lead qualification and proposal
20:49 generation system for a business. In the first tutorial, we'll be setting up the
20:53 base using AI to automatically qualify leads after they submit a form on their
20:56 website. In the second tutorial, we'll improve that qualification ability by
21:00 implementing an AI voice agent that can call the lead for more information. And
21:03 finally, in the third tutorial, we'll implement an automated proposal
21:06 generator that can instantly create proposals for qualified leads using the
21:09 information collected on the phone call and in the lead form. The whole point of
21:12 this is that as soon as someone's interested as contacting the business,
21:15 they can kick things off with them immediately by getting a proposal in
21:17 their hand. So this process of qualification is a crucial part of
21:21 running any business at scale because at the end of the day, not all people who
21:24 come to a business are going to be a good fit for their services. So for an
21:27 example, someone may run an accountancy business, but they only choose to work
21:30 with doctors. They are specialized in helping doctors with their finances. But
21:33 if a builder fills out their website form, then that lead would not be
21:37 qualified, right? Taking a call with them would be a waste of time because
21:40 they are not a doctor. Therefore, to stop wasting time, almost all businesses
21:43 need some kind of qualification system. And for qualified leads, most businesses
21:47 need to make some kind of proposal for them in order to kind of look at it and
21:50 and see what they're proposing and then agree to that proposal. Since most
21:54 proposals don't lead to a deal, maybe 20% if you're lucky, this is a huge
21:57 waste of time and resources for the company. So, long story short, what
22:00 we're about to build will solve a number of key problems for basically any
22:03 business, making it extremely valuable for you to be learning how to do it. And
22:05 in fact, this is something you'll be able to go and sell directly to
22:08 businesses when you're done. Here's what this process looks like without AI
22:11 automation first. So, a human sales rep must constantly check for new form
22:14 submissions from the website, review each lead's details, and evaluate if
22:18 they're worth pursuing, research the company to understand how to help them,
22:21 make phone calls, and deliver competent pitches, and then manually create custom
22:25 proposals. That is hours of repetitive work per week. And it could lead to
22:27 potential leads slipping through the cracks. If maybe they get the
22:30 qualification wrong, they don't understand what the business is doing,
22:32 they don't properly research it, or they're just too slow at getting back.
22:35 They may have to wait a whole day for the lead to hear anything back. So
22:37 decreasing the time it's going to take for a qualified lead to hear back from
22:41 them. It's going to drastically increase their conversions. With AI automation,
22:44 we can transform this tedious process into something much more efficient,
22:46 which is exactly what we're going to be building together. So here's how it
22:49 works. It's going to start as usual with a lead filling out the form on the
22:52 website. Then the system is going to automatically qualify them using AI.
22:56 Then it's going to further research the company using AI also. Then it's going
23:00 to send an automated phone call to pitch your offer using an outbound AI voice
23:04 agent. The system then saves the call outcome, summarizes the conversation,
23:07 and generates a personalized proposal for them, all without a human having to
23:11 lift a single finger and in a super scalable way. This is the power of AI
23:15 automation. We're taking tasks that used to demand hours of manual work and
23:18 turning them into workflows that run all by themselves. and we're going to build
23:21 it all using several of the most popular tools from the automation ecosystem. So,
23:24 if you're as excited to learn this as I'm to teach you, then let's get
23:27 started. And everything that you need to follow along with it, including all
23:30 resources, templates, prompts, etc., is going to be available for free on my
23:33 school community. You can find it in the first link in the description. You'll
23:36 need to request to join. It will take 1 to 2 minutes to be accepted. Once you're
23:39 in, you can just search for the title of this video, and then you'll be able to
23:42 find all the resources attached to it. So, that is how you can get all the
23:44 resources to follow along. Let's get stuck into it. All righty, guys. Just to clarify things
23:49 before we jump in, it's very important that you understand what we're doing in
23:52 this first build, but also in the second and the third and how it all fits
23:55 together and you get kind of an idea of where we're going. Understanding this
23:57 technology and then being thrown in a whole bunch of random things uh like Air
24:01 Table and and Slack and make and all of these different terms can be a bit
24:04 confusing. So, I just want very quickly to give you a bit of an orientation
24:07 before we jump into things of what we're building and why and how it works
24:09 because I've touched on it before, but I don't think I did a good enough job. So,
24:12 I'm filming this after just to really make sure you guys are fully clear on
24:15 what we're building here. So, and why it's valuable really as well. So um as
24:20 we touched on this is going to be an AI qualification system um for inbound
24:24 leads to a business. So easiest way to explain this is if we had a a website
24:28 and say we had this book a free consultation form and we might be
24:31 running ads to it. We might just have that on a website and people just
24:34 discover it by searching the web or or we're showing up in Google results. But
24:37 at some point, a business like the example I gave before, if you're an
24:41 accountant and you only work with uh builders or plumbers or whatever the the
24:44 example I gave, some people are going to fill out this form who are not
24:47 necessarily the right person for you. Like the my accountancy firm that only
24:51 works with plumbers, if a surgeon comes in or if a a pilot comes in and fills
24:54 this thing out and wants to get accounty services, we would ask them questions
24:57 around, okay, well, what's your name? What what do you do? Uh what sort of
25:00 industry do you work in? And we'll collect this information here. And as I
25:04 mentioned before, not all of the people who fill out your forms, particular if
25:07 you're running ads or or doing content and it's driving traffic to this landing
25:10 page as a business, not all of them are going to be people that you actually
25:13 want to get on calls with because if you get on calls with them and they're not
25:16 your ICP or the person that you could potentially work with, then it's going
25:19 to be a complete waste of time. So, a qualification is a very essential part
25:22 in in pretty much any business. And what you're going to build is a very very
25:24 powerful system starting off with something basic like this. And then
25:27 we're going to progress it to something a bit more advanced like this. And then
25:30 finally, something extremely valuable and powerful that you guys are all going
25:33 to be able to build along and ultimately sell this to someone if you wanted to
25:35 afterwards. So that's going to be covered in the monetization section. But
25:39 just to clarify sort of what we're going to be doing here in this first build, uh
25:43 you'll start off with building a form on a telly. So we're essentially
25:45 replicating this kind of website form, but we're just going to build it in
25:48 tally in this case. You can embed these into different websites, etc. But just
25:51 think of this as a form that people will get sent to. So maybe they clicked on an
25:54 ad and then they get sent to this. How can we help? We're going to build out
25:57 this form and then when people click submit on this form, it's going to send
26:00 them into this basically a fancy spreadsheet. And in this case, we're
26:03 using Air Table. Air Table is like a really awesome platform which you're
26:05 going to get to to use a lot within this build. Uh but Air Table, you can think
26:09 of it like a fancy spreadsheet and we're going to be able to map uh each of these
26:13 answers to a row in this database here. So every time someone fills this form
26:16 out that we create, it's going to load it and add it as a row into this
26:20 database. And the cool thing there is that air tableable allows us to within
26:24 make.com our workflow builder it allows us to set up events that when things
26:27 happen in air table like okay let's watch for new records that arrive and
26:31 that's what this here is doing. So this trigger of our automation is going to be
26:35 watching for new records that arrive in this and then it's going to take
26:39 anything that arrives in here say this first row and then it's going to put it
26:42 through this automation here. And so this is going to be a very basic one
26:45 where we've taken actually the AI component out of this. Before Air Table
26:48 really got advanced with the AI features. What you're doing here is take
26:51 the information from the Air Table. You'd use some chat GPT step here and
26:54 then you'd be able to do things afterwards. But in this case, I'm making
26:57 it extra simple for you guys and that we're actually just using Air Table's
27:00 built-in AI features. So down here, you can see in the tutorial that you're
27:03 about to go into, we will use Air Table to create an AI column here that's going
27:07 to take in information like uh like here where they're talking about their their
27:10 company, their budget, and about their needs. We can write a prompt in here and
27:13 it's going to automatically fill out this row. So as soon as a new person
27:17 arrives in this spreadsheet, this field that we're creating here is going to
27:20 automatically qualify them and use AI and this prompt that we write here to
27:23 analyze the information that they gave us and then output if they are qualified
27:27 or if they're not. So we're doing the qualification with AI built into Air
27:30 Table which is going to save us a lot of time and messing around with jumping
27:33 back and forth. So So that is basically the start of build one. We will do
27:37 immediate qualification with AI within Air Table and then we're going to if
27:40 they are qualified we're going to send an email to them and because we're
27:43 already doing the qualification here in Air Table it makes it a lot easier for
27:46 us over at make.com because for this particular node as you'll see in a
27:49 second we're going to set it up so that it's only actually going to look for uh
27:54 the rows or records in this table that have been AI qualified as qualified. So
27:58 if this prompt here that we've written analyzes all the information in the row
28:02 and then outputs qualified then and only then will it pass it into here. And this
28:05 is when we're going to be sending an email to the prospect and saying, "Hey,
28:09 we're interested in in hopping on a call. Here's a link to book in a call
28:12 with us." But at the same time, it'll also send a message to our sales team in
28:15 Slack and say, "Hey, look, we've just had a new qualified lead." So, this is
28:19 very much a a minimized version of what you could do uh maybe a year ago. But
28:24 because of the Air Tables AI features, uh we can do a lot of the heavy lifting
28:27 here and writing a prompt that's going to analyze what the person has filled
28:30 out here, aka are you a are you a builder or are you a pilot? And if they
28:34 were a builder, in the case of the example I just gave, it's we're going to
28:36 be using a different example in the build. Um, but it would come out as
28:39 saying qualified here. And because it says qualified, then it would
28:43 automatically trigger uh the rest of this automation, which is sending an
28:45 email to them and taking the next step and saying, "Yes, we are interested in
28:48 talking to you further, making sure that we're not talking to anyone that we
28:51 shouldn't be talking to and at the same time letting the sales team know." So,
28:54 that is build one. Um, that is how it all fits together. And I'm going to be
28:57 doing these little updates just before each build so that you guys are 100%
29:05 To get started, we need a way to gather our leads information. While there are
29:09 several great options for this build, we'll be using Tally, an easy to use
29:13 form builder. Just sign up for free or Then we'll click on new form and give
29:20 our form a title. Anyone using the form will see this title. So, we want it to
29:24 make sense for the user. So, in our case, we can name it, how can we help? I
29:27 want to build it from scratch. So, I'll hit enter. The way we create forms in
29:31 tally is with building blocks. So, if we click the plus button, we can select the
29:35 type of block we need from this list of input options. We'll start with a short
29:39 answer input for the user to write in their first name. And we'll do the same
29:43 for their last name. Notice how we're adding a label above each input, which
29:46 tells the user what to write in each of these. Then, we'll select the email
29:51 block for the user's email. Add in a phone number block. Then, another short answer input for the
29:57 name of their company. Since their budget is a number, we'll add a number
30:01 block to grab that. And finally, we'll add a long answer input where leads can
30:05 describe their specific needs. Now that our form inputs are ready, we can
30:08 customize the form's appearance to match our branding. We can adjust elements
30:14 like the background color, text color, button color, and the accent color. Feel free to
30:21 choose whichever colors you'd like. Now, we can go ahead and publish the form.
30:24 When it's ready, we'll see a sharable link. Let's copy this and actually visit the
30:30 published form. It's looking great and ready to use. In the real world, you
30:33 could either use this as a landing page or embed it into a website. Whenever a
30:38 new lead fills out this form, we need to be storing that response somewhere. For
30:42 our build, we'll store our form response inside Air Table. You can think of Air
30:45 Table kind of like Google Sheets on steroids. It's your database that not
30:50 only stores information, but can process information and if you push it far
30:53 enough, can even be used to build out more complex apps. So sign up for free
30:57 or sign in if you already have an account. On the home screen, you can see
31:02 that we could use AI to build things or start with a template like one for a
31:06 marketing campaign or a project tracker. But for our build, we'll start by
31:10 creating a base from scratch. Base is just Air Table's name for a database.
31:13 Again, it's really similar to a Google sheet or Excel sheet, but can be layered
31:18 with complexity. We'll name it lead base. We'll reference this name later
31:21 when setting up the connections between our tally form and air table and later
31:25 between make and air table. For every field in our form, we need a matching
31:29 field in air table. So, we'll add fields company, budget will reference this to determine
31:56 a lead's qualification. notes, details they submit about their needs created on
32:01 when the lead was added to this base. Importantly, we need to know if our lead
32:05 is qualified or not. Ultimately, lead qualification can be more nuanced, but
32:09 for our needs, we'll qualify leads whose budget is 10,000 plus. By using Air
32:13 Table's AI assistant, we can prompt it to create a custom field for us by
32:17 asking it to create an AI field called qualification that sets the lead as
32:23 qualified or not qualified based on whether the budget field is greater than
32:27 or equal to $10,000. Now, Air Table will use the power of an LLM open AI at the
32:31 time of this recording to help with creating this custom field. After
32:35 processing, the AI fields modal will appear. Since we want leads to be
32:38 autoqualified, we'll enable the automatic generation option.
32:44 If you encounter errors about missing or invalid fields while saving your AI
32:47 prompt, it's typically because the system can't clearly identify which
32:50 fields you're referencing. Double check that all fields are valid and properly
32:56 linked to the correct data in your base. We want our database not just to
33:00 autoqualify leads intelligently, but also to generate a descriptive message
33:04 about each lead that we can share with our sales team. So, we'll use the AI
33:08 assistant again to create a field that generates this message based on the
33:12 lead's information. After double-checking that this new prompt looks good, we can save it and
33:18 move on to the final field. We'll add a date field called contacted on that will
33:23 store the date when we contacted the lead through our automated email. In the
33:26 make workflow, we'll add the date to the field from make at the very end of our
33:30 workflow. Finally, we'll clean up this base by renaming the sheet to lead
33:35 contacts and set up each record or row to be called a lead. Now that we have
33:39 all the necessary Air Table fields that match our form fields, we can connect
33:43 Tally to Air Table via Tally's Air Table Just name the connection, select the
33:52 database and table you're syncing to, and map each of the tally fields to
33:55 their respective data fields in Air After saving this connection, we'll see
34:04 a confirmation that the integration was successful. Now, let's test the
34:07 connection by filling out the form with some dummy data, making sure to set the budget to $10,000
34:19 When we submit, we should see the new lead record populate in our Air Table
34:22 database. And voila, it's showing up perfectly. Since the budget is over
34:26 $10,000, it's automatically marked as qualified. And we've got an AI generated
34:30 message ready to send to our sales team via make. With our lead form synced up
34:33 to our database, we're ready to start building out the automation workflow.
34:37 This is what we're about to build out. We'll have a module that watches our Air
34:41 Table database for new leads. And when it finds one, it sends them an email via
34:45 the Gmail module, then updates that same leads air table row with the time we
34:49 contacted them. It also formats and sends a message into Slack channel. If
34:53 you're not familiar with Slack, it's a professional group chat with channels
34:56 dedicated to specific topics like marketing. With this plan in mind, let's
35:00 head over to make.com. Remember, this is our workflow builder,
35:03 the assembly line where we'll be constructing our automation. You can get
35:07 started for free or login if you already have an account. Once inside, you'll see
35:11 your dashboard, which displays info about your workflows, which are called
35:14 scenarios in Make. You'll see how busy they've been and how much data they've
35:18 used. This helps you monitor your make plan usage, which is particularly
35:21 important if you're on a free plan with usage limits. You can also see if any of
35:25 your scenarios require attention due to things like bugs and errors. The
35:29 scenario tab is where you'll spend most of your time, and you can organize your
35:32 scenarios into folders like you see here. Before we start creating our own,
35:35 I want to bring your attention to the fact that you have a bunch of templates
35:39 to get started from and adapt to your own use cases. And over time, you'll be
35:43 managing the connections here in this tab, which shows you which external apps
35:48 like OpenAI and Google you have connected to from within make. With that
35:52 brief tour out of the way, let's get into it and create a new scenario. We'll
35:56 start by setting up the trigger. After selecting the Air Table module, we'll
36:02 configure it to watch records. Since make doesn't yet know which
36:05 records to watch, we'll need to establish a connection between Make and
36:09 our Air Table account. We'll do so with a token because as you see in this
36:14 warning keys are deprecated. So we just need to create a token and paste it
36:18 here. To get that token go to your Air Table account settings,
36:28 and create a new token named something descriptive like make token.
36:33 When configuring the scope, set it to allow both reading and writing records
36:37 with full database access for read and I'm granting access to all current and
36:45 future bases, though you may want to restrict this depending on your specific
36:49 needs. Now, click create token and immediately copy and store it somewhere
36:52 secure. You'll never be able to see the complete token again. Once we paste the
36:56 token back into make, our connection to air table should be successfully synced
37:01 up. Now we just select the bas and table and set the trigger field to created on.
37:10 The trigger field tells make which field to monitor for changes in order to
37:13 determine when a record is new. By setting it to created on make will only
37:18 pull records whose timestamp is later than the last time it checked for new
37:22 records. That way it only acts on newly added leads and doesn't rerun on old
37:27 ones. As for the label field, we'll set it to the company. This field is just
37:30 for display purposes inside makes scenario logs. It helps you identify the
37:35 records that are being pulled in, but it doesn't change how the automation works.
37:39 The formula field lets you write an air table style formula that filters records
37:43 before they're passed into your automation. Make only processes records
37:47 when this formula evaluates to true. So, if a lead's qualification field is equal
37:52 to qualified, they'll be pushed through to the next step of the automation. If
37:55 they're not qualified, they don't pass this step. If you ever need help writing
37:59 these filters, you can ask an LLM like chat GPT or even have a chat with Make's
38:04 built-in AI. Once we hit save, we'll choose to start from now on. That way,
38:08 make only looks at new leads created after this point and skips anything
38:11 already sitting in Air Table. If we hit the run button, it should successfully
38:15 find the newly created qualified lead and display that records details. Once
38:19 you run this scenario, any qualified leads that were found will not be found
38:23 again on rerun due to this created onfield. You'll have to add a new lead
38:26 each time you want to test this scenario. Again, the automation is
38:31 filtering for only new leads. Since we want to send qualified leads an email to
38:35 schedule a call and notify our sales team about this lead via Slack, we'll
38:39 need to set up two actions. This is where a router comes in. It allows our
38:42 scenario to branch into these two separate tasks simultaneously. Without a
38:47 router, setting up tasks sequentially means if the email step fails, the Slack
38:51 step won't run at all. We want to avoid these waterfall effects where one
38:55 failure blocks other actions. Using a router prevents this issue while making
38:59 our scenario easier to troubleshoot and expand over time. On the first route,
39:03 we'll add the Gmail module and select the send an email action, which requires
39:08 us to set up a connection to our Gmail. If you're using a non-personal email
39:11 like mike@edge.ai, this setup is very simple. You just log into your Gmail account here. But if you
39:18 need to use a personal Gmail, one that actually ends in gmail.com, the process
39:22 is more involved. So, I'm going to ask you to be patient because there are
39:26 quite a few steps involved to connect a personal Gmail account here. If you're
39:29 just watching and not building along yet, feel free to jump past this
39:33 section. Same goes for those of you who are using a non-aggmail.com email. But
39:37 if you do need to connect your personal Gmail account, here's what that looks
39:40 like. As you can see, if we pop open these advanced settings, this requires
39:45 two key pieces of information. a client ID and a client secret. These act as the
39:49 key that allows make to unlock access to your Gmail account. Make provides
39:53 instructions for how to generate these, which you can find by clicking on this
39:56 guide link. At the time of this recording, it links to the help center.
39:59 From here, click into apps documentation, then click on communication and scroll down to Gmail.
40:05 If you're wondering why make didn't just link us to this page, so am I. From
40:09 here, we'll click on create a custom oorthth client. And now we need to log
40:14 into the Google Cloud Platform. Once you log into Google Cloud with the Google
40:17 account you want to send emails from, you'll click select project at the top
40:21 left. From there, click new project, then give the project a name like make and hit
40:29 create. Once it's created, a notification will appear in the top right corner. Click
40:35 select project to open your new project. Now that we're inside, let's enable the
40:41 Gmail API. Go to APIs and services and click enable APIs and services. In
40:45 the search bar, type Gmail and select Then click enable. Next, we need to tell Google who is
40:57 requesting access and what kind of data is being requested. So, we'll click on
41:01 Oorth consent screen and hit get started. To fill out the app
41:05 information, we need to give the app its name. So to keep things consistent,
41:09 let's call it make like we did for the project name earlier. We also need to
41:12 tell it which email we want to give these new abilities to. Now we select
41:17 external for the audience type and next. Under contact information,
41:22 enter an email for Google to notify you about this project. Hit next. Then agree
41:27 to the terms, continue, and hit create. We now need to give make permission to
41:31 interact with Gmail on your behalf. So head into the data access tab and click
41:36 add or remove scopes. We're adding Gmail API scopes. So, search for Gmail and
41:41 then select the following scopes. Both scopes to read, compose, and send
41:46 emails. Manage drafts and send view your emails and settings. View your metadata.
41:51 Add emails to your mailbox. Send email on your behalf. See, and edit labels.
41:55 With all those labels selected, hit update. Then, scroll down and make sure
41:59 to save these scopes. Next, let's configure the branding settings. Scroll
42:03 down to authorize domains and add both make.com and integromat.com which is the
42:08 old name make used to be called. Then save. After that head to the audience
42:13 section and click add users where you'll add your Gmail address and
42:18 hit save. Now it's finally time to create the actual credentials that make will use.
42:23 Go to the client section and click create client. For the application type, choose web
42:30 application and name it make. Scroll down to authorize redirect URIs.
42:35 Click add URIs and paste the exact redirect URI provided by Make. You can
42:39 find this in their Gmail integration Click create and you'll be given your
42:53 You will copy both of these and paste them into the field in makes Gmail
43:01 Now we can click sign in with Google. This sign-in window will appear. Select
43:05 your Gmail account. Grant the requested permissions and hit continue. Once
43:09 that's done, your Gmail account should be successfully connected to make. I
43:13 know that was a lot and I wish it was simpler too. But the good news is we can
43:17 now start sending emails from that Gmail account. So, let's now create the email
43:21 template. We'll set up a subject line using the lead's name from their Air
43:25 Table record and whatever engaging text you want to add in here. For the content
43:29 of the email itself, we can address them by name and say, "We received the
43:32 request about their company and would love to discuss their goals on a call
43:38 they can book via our scheduling link. For this, we will be using a handy tool
43:42 called Kalanley. If you're not familiar, Calendarly is a scheduling tool that
43:45 lets people book meetings with you. You connect it to your calendar, set your
43:49 availability, add conferencing tools like Zoom or Google Meet, then create
43:53 event types like an introductory call and share a link for people to book
43:56 these meetings with you. So just set up a new account if you don't have one or
44:02 log in and we'll create a new event type and call it intro call. This event will
44:06 have a 30-inut duration and the location will be Google Meet. Keep in mind you
44:09 may have to connect a conferencing tool if you're using something like Zoom. Now
44:13 we can set our availability for this intro call event. I already have mine
44:17 set up, but to do that, you just head to the availability tab and go to calendar
44:21 settings. Here you can see I already have a couple calendars synced up, but
44:25 to add a new one, click connect to calendar account and select your
44:28 provider, such as Google. Then just choose the account you want to sync and
44:32 log in. Again, I've already set this up, so I'll close out of this and head back
44:37 to the event. For each event, we can set the availability for that event type. I
44:40 may want to change the time I'm free on Friday, either a specific day or every
44:45 Friday. When I apply these changes, that will update the availability for just
44:49 this event type and does not apply to other events I may have set up in
44:52 Calendarly. Down in more options, I can do things like add an event description
44:56 where I can tell the person booking that I'm excited for our chat. Once we hit
45:00 save changes, the event is ready to share. So, let's copy this link and head
45:05 back into make and paste it into the body of the email we're sending to our
45:08 lead. Finally, we of course need to add the recipient of this email. So, we'll
45:11 fill that in with our lead's email address. With that set up, we can move
45:15 on to our other route below. Ultimately, we're going to be posting a message to a
45:19 Slack channel about our new lead. But if you don't use Slack and don't want to go
45:22 through the steps of connecting another external tool, you could sub this out
45:26 for another Gmail module where we send an email to our sales team letting them
45:29 know that we have a new qualified lead and give them the lead's name and other
45:34 relevant info. Then set the recipient email to whichever test email here. But since I
45:40 want to show you how to use a bunch of popular tools, let's look at setting
45:43 this up to post to Slack. We'll choose the Slack app and select the create a
45:48 message action. If you already have a Slack organization, you can simply
45:56 if you don't, let's quickly walk through creating a new one. You'll head to and
46:00 sign in with a new account, then click create a workspace. Give it a name.
46:04 Here, I'm just adding the name of our mock agency that we're qualifying leads
46:07 for. Then just run through these steps to confirm your name. Skip adding new
46:11 members for now and start with the free version. Now we're in our new Slack
46:15 organization. As you can see, there are channels here for different topics and
46:19 direct messages down here. We'll add a new channel called marketing because
46:22 this is where we'll be posting our Slack message into for the entire sales team
46:26 of our imaginary agency to see. With these ready to go, we can head back to
46:30 make and connect the Slack module to our We'll tell the module where to post the
46:38 message by selecting from a list of our public channels, specifying the
46:42 marketing channel we just created. Now, what should it post to this channel? It
46:45 should post the AI message from our Air Table lead. So, we'll add that in here.
46:49 If we run the scenario, it will work, but the message will look messy because
46:53 it's technically a collection, which just means it includes extra stuff that
46:57 we don't want or need to display here. We can solve this by inserting a
47:00 formatter module just before we post to Slack. We'll select a text passer, which
47:05 will do a pattern match using a regular expression to pull just the clean
47:08 message text out of it. While this might look complicated, think of it like a
47:12 smart highlighter that scans the text and grabs only the part we care about,
47:15 the messages value itself and not the messy metadata. If you ever need to use
47:19 a regular expression or rejects in the future, just ask an LLM to draft one for
47:24 your use case. Now, we tell our text passer which text to pass, the AI
47:28 message. With that formatter taken care of, we can head back into the Slack app
47:32 and tell it to create its message based off that freshly passed text instead of
47:38 the original messy version. As a final step, we want to circle back
47:42 to Air Table and update our lead record with information about when we auto
47:46 contacted them. After the Gmail module, we'll add an Air Table module and use
47:50 the update a record action. We'll configure it to return to the same lead
47:54 base and contacts table targeting the ID of the record that just pushed through
47:57 this scenario. We'll set the contacted onfield to the current date and time by using the now
48:03 expression which simply tells make to insert the exact moment the automation
48:07 runs. If we run the scenario and head back into air table, we'll see the date
48:11 was successfully added to the contacted on field giving our sales team context
48:16 about when the lead was auto that email. Our scenario is all set up and working
48:20 great, but up until now, we've only been running it manually by clicking the run
48:24 once button by toggling on the schedule. We can configure our scenario to run on
48:29 autopilot at regular intervals such as every 15 minutes or on a specific day of
48:33 the week or month. You can even set a custom schedule using advanced
48:36 scheduling using time ranges with start and end dates if you only want it to run
48:40 during a certain window. Once you're confident it's working, just flip that
48:44 schedule toggle on and your automation will run in the background while you
48:47 focus on more important things. Of course, you'll want to frequently save
48:51 the scenario. And if you ever need to revert to a previous version, you can
48:55 revert to that version from here. Scenario inputs are useful for more
48:59 advanced use cases, like when one scenario's output becomes another
49:03 scenario's input, but that's beyond the scope of this basic build. You've also
49:06 got your scenario settings, a place for notes, an auto aligner if your workspace
49:11 gets a little messy, and even a little Finally, there's a quick reference to
49:19 every app and module used in your scenario, helpful for getting a bird's
49:23 eye view of your automation. And so, that's build one explained, just a very
49:27 basic AI qualification based out of Air Table. But to give you an idea of where
49:29 we're going with this, in the next build, build two, we're going to be
49:32 adding in a voice agent here. You may have heard about voice agents before,
49:35 but they're are a really a really exciting uh area of the AI space right
49:38 now. You have two main types. You have inbound and outbound. And uh inbound is
49:42 when you can set up a phone number and people can call. When people call that
49:46 number, then they get to talk to the AI directly. Outbound voice agents is what
49:50 we're going to be building here where we can initiate a call using our automation
49:55 here uh to send a call out to someone and say my phone starts ringing and I
49:58 can pick it up and I'm talking to the AI that we've created here. So this is
50:01 really really cool stuff and super powerful. But the reason we're doing
50:04 this is because the as you will have seen on the form that we set up, we have
50:08 quite a limited amount of information that we're collecting from the business
50:10 and this is for good reason. You don't want to put too much information here uh
50:14 or it will decrease the number of people who fill it out. So sort of having a
50:17 lean form and then if we initially qualify them here in the second build
50:20 we're going to expand it so that we can actually once they are qualified as
50:24 we've explained here the AI and air table is going to determine
50:27 preliminarily if they are qualified then we will use our voice agent to actually
50:31 call the person and ask for more information and we'll walk through a
50:34 script basically and ask them hey tell me a bit about your business and your
50:36 needs what are you hoping to get out of us da da da da basically collecting a
50:40 bunch more information that can then be used for even greater and more accurate
50:44 qualification. So here we have our Vappy voice agent which is going to call them
50:47 and then we're going to get the data of that call, the transcription of that
50:50 call and Vap is actually going to be able to analyze it for us to determine
50:54 if it was a successful qualification or not. And then we have to build the
50:56 automation to handle a few different cases because of course not everyone's
50:59 going to pick up the phone. So here we create a route that handles if they
51:02 answered the phone and then based off the information were they interested or
51:05 were they not interested and if they didn't answer we have some other things
51:08 that we can do to handle it here and update the air table and things like
51:11 that. So, it's essentially build two is adding on top of what we've already
51:15 built on build one uh where we have just a basic qualification and sort of send
51:19 them an email to book in a call or to let the sales team know. Here, we're
51:23 trying to do even deeper qualification and making sure that our sales team
51:26 really isn't getting on any calls they shouldn't be by sending out a voice
51:29 agent to collect more information for us to do an AI pre-qualification for this
51:32 business's lead. So, let's take this automation to the next level by adding
51:36 AI voice calls to qualify leads even more effectively. We've already built a
51:39 strong foundation, a system that identifies qualified leads, follows up
51:44 via email, and notifies your team. This alone puts you ahead of the curve. But
51:47 what if we could make the system even more responsive, more conversational,
51:51 more human even? That's where an AI voice agent comes in. By weaving it into
51:55 our workflow, we can automatically call leads, gauge their interest through real
51:58 conversation, and tailor our follow-up based on their responses, all without
52:02 lifting a finger. This is where automation becomes more than a task
52:05 runner. It becomes a teammate. It's the closest thing we have to a human
52:08 conversation at scale. And when it comes to leads, timing and tone matter.
52:12 Research shows that responding to leads immediately increases success rates by 7
52:16 to 9x. In this next section, we'll integrate VP, a popular AI voice agent
52:20 service, into our workflow. First, we'll set up our voice agent. Then, we'll
52:23 configure make to have it call our leads. Finally, we'll enhance its
52:26 effectiveness by feeding it custom research about each lead, enabling more
52:30 targeted pitches. So, what exactly is an AI powered voice agent? You can think of
52:33 them like a version of Siri or Alexa, but one that's specifically designed to
52:37 have natural phone conversations with people. Depending on your needs, it can
52:40 either make calls for you or answer calls. These agents can do amazing
52:43 things like follow up with potential customers, schedule appointments, answer
52:47 common questions, collect feedback, or conduct surveys. There are several
52:50 platforms that offer AI voice agents like 11 Labs, great for highly realistic
52:55 voices, and Vappy, known for being fast and affordable for our automation. We'll
52:58 be using Vapy because it's beginner friendly, cost-ffective, and integrates
53:02 well with other tools. If you haven't already created an account, go ahead and
53:06 do that first. Once inside, we'll go to the assistance tab and click create assistant. This
53:12 starts the process of configuring our assistant. You can start from a template
53:15 like a lead qualification specialist, but we'll create our own from scratch
53:21 and name it Ben. When choosing which provider and model to use, you'll want to balance ability
53:26 with speed and cost. So, keep the price in mind and the time it takes to
53:30 actually respond, which we call latency, which you can see next to the model
53:33 options. For now, we'll go with the cheapest and fastest option, which is
53:37 GPT4.1 Mini at the time of this recording. The most crucial element is
53:41 the prompt, which defines your agents behavior and objectives during the call.
53:45 First, we specify the opening greeting message that the agent will use when
53:49 calling someone. We'll open up the call asking if it's a good time to talk about
53:53 the lead's business needs. The actual prompt is much more detailed
53:58 and I'll paste in one I wrote earlier. You can find this prompt and everything
54:01 else you'll need to follow along in the first link in the video description.
54:06 Let's take a look at what it includes. As you can see, we start off by defining
54:11 its identity and role. A voice assistant representing Edge AI and AI automation
54:16 agency. then specify its goal to pitch our services and determine whether the lead
54:22 would like to receive a proposal. We then set its tone and behavior. A
54:27 knowledgeable rep who is curious but not pushy and who uses casual natural
54:32 language with words like uh and mhm. This section helps humanize your
54:36 assistant. Finally, we outline the structure of the call that it should
54:40 replicate. The conversation flow starts with introducing our AI services, then
54:44 moves to understanding the prospect's needs through targeted questions. After
54:48 acknowledging their responses, we pivot to offering a proposal. If pricing comes
54:53 up, we defer to the proposal for details. The call ends either by
54:56 confirming they'll send the proposal over email or with a plight goodbye if
55:00 they decline. It's best to outline the steps as clear bullet points like this,
55:04 but keep them focused. Too many steps can overwhelm the AI assistant. In our
55:08 prompt, we included context about the company the assistant is representing.
55:11 For our needs, this is sufficient. But if our call was more involved, like a
55:14 customer support agent who is receiving hundreds of nuanced calls, we could add
55:18 files into its knowledge base that it can reference, like all of our policies
55:22 and procedures. But keep in mind, this will increase the response latency since
55:25 the assistant will need to make round trips to these files as it generates its
55:29 responses. So, if you can get away without adding supplemental files, your
55:32 agent will be more performant. Our assistant is almost ready to start
55:35 making calls on our behalf. But before we publish it, let's add some extra
55:39 configuration. As we scroll down, we'll see that we can tweak the transcriber
55:42 settings, which we'll leave alone for now. We can also select different voices
55:46 depending on the personality we want our agent to have. For our agent, we'll go
55:49 with Vap's Elliot voice, but feel free to try out different options to feel out
55:53 which one is best for your use case. You can also add background noise to the
55:57 call. Adding ambient sound makes conversations feel more natural and
56:01 helps mask any brief delays that occur while the agent generates responses
56:04 instead of awkward silence. Callers will hear realistic background sounds. Next,
56:08 you'll see the tools section, which unlocks powerful ways for your agent to
56:12 take action during a call, like sending data to your CRM, triggering external
56:17 workflows, or running custom logic. This is where Vappy really shines for more
56:21 advanced automations. However, since it's beyond the scope of this tutorial,
56:24 we'll leave it untouched for now. What we will be using is the call summary.
56:28 Vari automatically transcribes the call and the prompt inside here tells it how
56:31 to summarize the call based on that transcript. If we open it up, we'll see
56:35 there's already a built-in prompt that tells the assistant how to create its
56:39 call summary. While this default prompt works decently, we can make it even more
56:42 valuable for our make scenario by customizing it to capture additional
56:46 details. Here, I'm pasting in a new prompt telling it to summarize the call
56:50 in a couple sentences with details that will help out with creating a proposal
56:53 later on. Since we'll add this call summary to our lead's air table record,
56:57 capturing detailed insights here is valuable. We also want to tell Vappy how
57:00 to judge if a call was successful or not. There's already a decent default
57:04 prompt here, but we'll customize it for our needs by telling it to judge success
57:08 solely on whether the lead wants to receive a proposal. We'll evaluate
57:11 success with a pass fail, which simply gives us a true or false value that we
57:14 can use in our automation workflow. While there are additional advanced
57:18 options here, the final step we're concerned about is the voicemail
57:22 detection. Let's make sure we toggle that on with VP as the provider. This is
57:26 helpful for our make scenario since we need to accurately track whether the
57:30 call was actually answered or if it just went to voicemail. We're now ready to
57:34 hit publish and test our newly created voice assistant. Hey, it's Ben from
57:39 Edji. Is now a good time to talk about your business needs? Yes, it is. Great.
57:44 I'm following up on your submission because we build custom AI agents and
57:48 automations for companies like yours. anything from customerf facing bots to
57:52 internal workflow optimization. I'd love to understand a bit more about what led
57:57 you uh to reach out. What sort of processes or tasks are you looking to
58:01 automate? I'm looking to implement custom AI agents to facilitate learning
58:05 on an e-learning platform. That sounds really interesting. Um so you're looking
58:10 um to enhance uh the learning experience on your e-learning platform with AI
58:14 agents. Are there any specific challenges or bottlenecks you're facing
58:18 right now that you hope these agents could help with? Uh, yeah, just
58:21 streamlining the learning of the content library on my platform.
58:26 Got it. So, you're looking to streamline how users access and learn from your
58:30 content library. That makes a lot of sense. Thanks for sharing that. Would it
58:34 be helpful if I sent a quick proposal your way showing how we could support
58:40 your goals? Yeah, that'd be great. Perfect. I'll send that to the email we
58:44 have on file. Thanks again for your time and I look forward to helping you with
58:49 your project. Great. Thanks. Bye. You're welcome. Have a great day. Bye. Great.
58:54 Our assistant is working as intended. It's now the voice of our lead outreach
58:58 standing by, ready to make calls on our behalf. Before integrating this
59:01 assistant inside of Make, let's walk through the new path a lead will take
59:04 through our extended scenario. We left off here, but we're going to extend this
59:08 for the trigger. Nothing changes here. A new Air Table record marked qualified
59:12 still kicks things off. Then our voice agent will call the lead using the
59:15 number from their record. We'll add a short pause to wait for the conversation
59:18 to finish before we analyze what happened. We grab the call record from
59:22 VAP. Then we check if the call was answered. Answered. Great. We log the
59:27 call summary and mark the lead as interested or not in Air Table. No
59:32 answer. No problem. We fall back to emailing them a link to schedule a call
59:35 and alert our sales team. Since we're going to be logging call summaries and
59:39 interest level about our leads, we'll just need to head to our lead base in
59:43 Air Table and add those. We'll add a long text field called summary,
59:48 a checkbox for interested, and while we're at it, let's add a new date for
59:52 when a proposal was sent on since in the final phase of this buildout, we'll be
59:55 generating and sending proposals to these interested leads. Back over in
59:59 make, we're ready to extend the lead qualifier scenario that we built
60:03 earlier. To keep things clean, let's just clone what we built and extend from
60:07 there. Name it lead qualifier plus voice agent. And now we're ready to continue
60:11 building. The beginning step remain the same where we watch for new leads.
60:17 Then we'll add the VP module to create an outbound phone call.
60:22 We will need to set up the connection to our VP account, which means we'll head back into that
60:28 dashboard, click on API keys, then copy your private key. If you don't see one
60:31 in there, just go ahead and add a new one. Give it a relevant name. Decide if
60:35 you want to restrict it to only work on certain sites. Or if you only want it to
60:39 work with specific assistant, then create the private token. You'll see it
60:43 pop up here. Then just copy it. But make sure to save this somewhere safe since
60:47 you won't be able to view it again. Pasting the key into the VP module
60:50 should set up our connection. Then we can configure our call. We'll just fill
60:53 in these fields, giving it the assistance ID, which you copy from the
60:58 top of the Vappy Assistant page. We provide the assistant with the lead's
61:01 phone number pulled from their air table record. But what number is our assistant
61:06 calling from? If we head back into Vappy and click on the phone numbers tab, we
61:09 can create a new free phone number with Vappy, specifying the area code to call
61:13 from. Note that at the time of this recording, only US area codes are
61:17 supported by Vappy. Since we're building this assistant for learning purposes, we
61:23 can simply use a US area code like 223. However, if you plan to deploy an
61:27 assistant for production use and you or your client are located outside the US,
61:30 you'll need to import a number from something like Twilio or Vonnage. Okay.
61:34 Once that number is created, we can copy its ID, not the number itself, and paste
61:38 it into the makevarpy module. And with that, our voice agent is set up and
61:42 ready to make calls triggered by our make scenario. Once the call is sent, we
61:45 need to wait a bit before we check what happened. Calls take time to connect,
61:49 ring, and potentially have a conversation. If we try to fetch the
61:52 result too quickly, we could just get an empty response or an error. So, let's
61:56 add a short pause. Insert a tools sleep module. Set the delay to 300 seconds.
62:01 This little breather, 5 minutes exactly, gives Vappy time to do its thing before
62:05 we peek in to see how it went. To ensure things are working so far, let's add a
62:08 new lead to our Air Table base. We could add them via the form, but it's quicker
62:12 just to duplicate an existing lead in order to create a new one. Now, we'll
62:15 head back and make and run the scenario Hey, it's Ben from Edgeai. Is now a good
62:25 time to talk about your business? Yeah, it is. Great. I'm following up on your
62:32 submission about exploring AI solutions. We build custom AI agents and
62:36 automations for companies like yours. Anything from customerf facing bots to
62:40 internal workflow optimization. I'd love to understand a bit more about what led
62:45 you to reach out. What sort of processes or tasks are you looking to automate?
62:49 Yeah. So, I'm just interested in um seeing how we can build out some maybe
62:53 AI powered workflows to help the That sounds interesting. AI powered
63:02 workflows can really streamline operations. Are there any specific
63:06 bottlenecks or manual tasks you're hoping to reduce? Hoping to help my uh
63:12 learning management system run more Got it. Optimizing your learning
63:19 management system can definitely enhance efficiency and user experience. Uh,
63:24 thanks for sharing that. Um, would it be helpful if I sent a quick proposal your
63:27 way showing how we could support your Perfect. I'll send that to the email we
63:36 have on file. Thanks again and I look forward to your feedback. Thank you.
63:41 You're welcome. Have a great day. Now comes the part where the automation
63:45 becomes observant. We're not just automating calls. We're building a
63:48 system that pays attention to what happened. Did someone pick up? Did the
63:52 assistant get the job done? Was the lead interested? To answer those questions,
63:56 we need to ask VBY for the results of a specific call. And to do that, we'll use
64:00 an HTTP module to make a request for that call record. To better understand
64:04 this module's function, let's explore what an HTTP request is and how it fits
64:08 into the broader workings of the internet and how APIs fit into all of
64:13 this. I know it sounds complicated, but let's break it down with a simple
64:16 analogy. Remember earlier we learned how an automation workflow is like a facto's
64:20 assembly line where each module is like a machine performing a step in the
64:24 process? Well, factories don't operate in isolation. Sometimes your factory
64:28 needs supplies, information, or services from outside its own walls. A real
64:32 factory might call a supplier to order more raw materials. Ask a logistics
64:37 company where a shipment is. Verify parts. Meet quality standards with lab
64:41 testing. request a maintenance crew to check the temperature of a remote
64:44 machine. Simply put, a factory coordinates with external partners to
64:49 handle tasks outside its own expertise. Just like real world suppliers don't
64:52 take factory orders shouted over a fence, external services need a
64:56 structured way to receive requests. That's what an API or application
65:00 programming interface is. In our factory analogy, the API is like the official
65:04 order form that your external partners use to process requests, which you would
65:07 fill out to order some special machine parts to use within your factory. Online
65:12 services like Vappy, Air Table, or Slack all offer APIs that follow the same
65:15 principle. They give outside entities like make.com a structured way to send
65:20 and receive information. APIs give us a clear, consistent way to ask for
65:23 something and get a predictable, reliable response. There are several
65:27 types of HTTP requests that serve different purposes. So, you could be
65:32 saying, "Get me this thing. Post or add this new thing. Put this info where it
65:37 belongs. Replace the whole thing. Patch just this one part of the info, but
65:41 don't replace the whole thing. Delete this thing." In our case, since we're
65:45 about to be making an HTTP get request inside of Make, we're essentially
65:49 saying, "Get me this call summary. Here's what I want, and here's who I am.
65:52 I've got the proper permission to access what I'm requesting." While platforms
65:57 like make.com and n10 provide a visual interface for building workflows, under
66:00 the hood, they're actually making API calls to connect with these external
66:03 tools. So even though you're working with visual blocks, these modules use
66:07 the same underlying language that developers use to connect services
66:11 across the web. Yes, this is no code development, but that doesn't mean code
66:13 isn't running. It's just happening behind the scenes. That was quite the
66:17 detour, but an important one because it gives you a firm grasp on how things
66:20 work on a deeper level. Now that you understand what an HTTP request is and
66:26 how it fits into the big picture, let's set this module up to go get the record
66:30 for the call that was just made. We'll set the method to get. But where are we
66:33 getting something from? Well, we're going to make a request to VP's API to
66:38 get the call that just happened. If we check out the VP documentation, we can
66:42 see that we need to send our request to this URL where the last part of the URL
66:47 is the ID of the call we're fetching. So let's copy that URL and paste it into
66:51 the HTTP module. Adding on the call ID from the vari modules output
66:56 in the header section. Here's where we add instructions like putting a label on
66:59 a package by putting authorization in the name field. This just means we're
67:03 saying we have the authorization to get this info and the proof is the value of
67:07 the header itself. Another way to think of that proof is that we have a key to
67:11 unlock this special box that contains the information we're requesting.
67:17 That key is the API key that we generate from VPY. So in the value field, we'll
67:21 type bearer. Then paste that key next to it. This roughly translates to the
67:26 person who possesses or bears. This key has permission to access what they are
67:30 requesting. Finally, we'll say yes to passing the response. This means the
67:34 module will break apart the response into structured fields that we can
67:38 easily use later in the scenario. Let's make sure this is working by running
67:41 this module only. Here at the bottom, it's asking for the call ID that we want
67:45 to get, which we can just grab off of the last time the VAP module made a
67:50 call. Placing that calls ID here and hitting save will now cause this HTTP
67:54 get request to fire off. And we quickly see that it was a success. We now have
67:57 access to all of this data from our call, including the summary, which as
68:01 you recall is the result of that prompt we added into VP to summarize the call
68:05 for a sales team. We also have access to the analysis, which tells us if the call
68:08 was successful. This is perfect since we'll use both pieces later in the
68:12 scenario to update the lead's air table record with their call summary and
68:15 interest level. So now that we're monitoring the call results, let's start
68:19 implementing those next steps. First, we want to determine whether the call was
68:22 answered. So we'll set up a router to create paths for both answered and
68:26 unanswered calls. When someone answers the call and engages with your voice
68:30 assistant, it's a valuable interaction that your sales team needs to know
68:33 about. Let's set up a filter to check if someone answered the call. On the first
68:37 route, we'll add a filter with a condition that checks for calls where
68:41 the data do the ended reason equals customer ended call. Among all
68:46 the data points available, this is our most reliable indicator that the call
68:50 was answered and didn't go to voicemail. On the second route, we'll do the
68:53 opposite and set the condition to filter for cases where data ended reason does
68:59 not equal customer ended call. On the answered route, we want to know
69:03 if the lead was interested in being sent a proposal or not. So we'll add another
69:06 router that splits into two paths based on the call analysis, specifically the
69:11 success evaluation. Remember earlier when we told Varie to evaluate if the
69:15 call was a success based on whether the lead wanted a proposal or not. That
69:18 evaluation is contained in this data here. Since this data is simply true or
69:22 false, we'll check whether the analysis is equal to true success or false
69:28 failure on each route. For both routes, we want to update the lead's record with
69:31 the call summary and note whether they were interested in a proposal. So on
69:35 each route we'll add an air table update a record module. On the interested route
69:39 we'll target the lead contacts based by this lead's ID and update the contacted
69:45 on field two. Now add the call summary We'll do the same for the not interested
69:58 but check no for interested. Now, our sales rep can instantly see when the
70:01 lead was contacted and what transpired during the call. The system
70:04 automatically documents each interaction, just as a human rep would
70:07 take notes after speaking with the prospect. By the way, in case you're
70:11 curious, this process technically performs a put request under the hood
70:15 where it puts or updates something about the record without replacing it all
70:19 together. Of course, not all of our VP calls will be answered. This happens,
70:23 people are busy or they screen unknown numbers. That's why we build in a
70:26 backup. On the notwans answered route, we'll keep the steps from our initial
70:29 build where we send that lead an email to schedule a call and we let our sales
70:33 team know about them. With this all locked in place, let's run the scenario
70:38 to make sure it's working. Hey, it's Ben from Edge AI. Is now a good time to
70:42 quickly talk about your business needs? Yes, it is. Great. Thank you. I'd love
70:47 to understand a bit more what led you to reach out to us. Are there specific
70:50 processes or tasks you're looking to automate? Yeah, looking to uh implement
70:55 AI agents. That's exciting. Implementing AI agents can really streamline operations. Um are
71:01 there particular tasks or workflows at your company that you find especially
71:04 timeconuming or manual which you're hoping to improve with automation? Yeah,
71:09 just hoping to um have the agents trained on the learning material so that
71:13 they can help my students. That makes a lot of sense. Training AI
71:16 agents to assist your students could really enhance their learning
71:18 experience. Thanks for sharing that. Would it be helpful if I sent a quick
71:23 proposal your way showing how we could support you in developing those AI
71:27 agents? Yeah, that would be perfect. Excellent. I'll send the proposal to the
71:34 email we have on file. Thanks again. We've done a lot. Let's zoom out for a
71:37 second to recap. With our current scenario, we now have a system that
71:41 detects qualified leads, places an AI powered voice call with no manual
71:45 effort, waits, listens, and reacts depending on whether someone answered
71:47 and whether they were interested or not, logging the outcome accordingly. As of
71:51 now, the voice assistant works, but it's not as capable as it could be. Without
71:55 knowing our lead's name or anything about their company and needs, the
71:59 agents ability to pitch is pretty limited. Our next step is to enhance our
72:03 scenario by researching our lead with open AI and passing those insights to
72:07 our Vappy Assistant, allowing it to personalize each call. Essentially,
72:11 we're going to have a a search feature that is going to not only just use the
72:15 voice agent for uh doing the research and getting more information, we're
72:19 going to use OpenAI's uh search models. So, when people fill out that form we
72:23 made, then we're going to use OpenAI to research the internet for that lead and
72:26 get some information on that. Then, we're going to send a call to them again
72:29 using Vappy. This time it's going to be personalized with the information that
72:32 we got from that web search. So it's really, hey, we know this about you, but
72:35 what else? We're looking for this information on you. You can really get a
72:39 a a complete picture of who this person is before we even booked them in for a
72:42 call with our sales team. So with the VIP module we're currently using,
72:46 unfortunately, we're not able to feed anything into it, at least not at the
72:49 time of this recording based on the VIP module's current setup, but we can solve
72:53 for this by switching out the VP module with a more custom approach using an
72:57 HTTP module. So, let's drag it out of the workflow and unlink it. And since
73:00 we'll be borrowing some of the values from it soon inside a new HTTP make a request module.
73:07 Since we're making a manual request out to the Vappy API, we need to specify the
73:12 URL just like we did in our existing HTTP module. So, we can go ahead and
73:18 copy that URL from the get module. Since we're simply placing a Vappy call and
73:22 not retrieving an existing one, we won't need the calls ID. So, we'll leave that
73:26 off. And we'll change the method type to but we need to add a second item where
73:37 the name is content type. Value is application/json. Content type is like labeling the
73:43 envelope you're sending. Application/json means inside this request, the data is structured like a
73:50 JSON object. It's like writing English or Spanish on the outside of a letter so
73:54 the recipient knows what language to expect when reading it. Continuing down
73:58 the module, we'll then set body type to raw, which means we're manually writing
74:02 out the data we want to send to VP. And we set content type to application/json,
74:09 which again tells VPY to expect data formatted as JSON. And I know many of
74:12 you don't know what JSON is, but it's less intimidating than you might think.
74:17 It's really just a pair, the key and the value. Just like in a spreadsheet where
74:21 you have the keys like phones, name, and email. Then you have the values, the
74:26 actual data that goes with each key. It's an easy to read way to structure
74:30 and share data. In JSON, everything sits inside curly brackets. This is called an
74:34 object. Both the keys and values need to be in quotes with commas separating each
74:38 key value pair. Don't worry if you're not a JSON expert. Many free online
74:43 tools can help you check if your JSON formatting is correct. So back in make
74:47 down in the body of our HTTP call out to Vappy, we're going to add some JSON to
74:51 save some time. We'll grab some values from the VPY module we were using
74:54 earlier, including the assistant ID, which we'll paste into the JSON.
74:59 And we'll also grab the VPY phone number ID off that old VP module and paste
75:05 reuse it in our new HTTP module. We're just doing it a bit more manually. The
75:08 main difference here is in this assistant override section. As it
75:12 sounds, we are overriding the assistant within custom variables. These are
75:15 essentially placeholders that will be replaced by the lead's name, company,
75:20 and the research we perform for each of them. This way, when we say, "Hey, Vapy,
75:24 call this lead." We're also saying, "And here's info about them to use on the
75:27 call." Now, that we're going to be sending these variables into the
75:31 assistant, we need to head back over to the Vappy dashboard and tell our
75:34 assistant Ben to be expecting that information and instruct them on how to
75:38 use it. In the first message, we can add the first name variable so our assistant
75:42 can greet our lead by name. In the updated prompt, we'll inform Ben that he
75:46 will receive custom data via variables such as the lead's first name, company,
75:50 and company research. And to use these details to personalize everything on the
75:54 call with the goal of pitching more effectively. With our assistant ready to
75:57 receive all that custom info, we just need to perform research on our lead.
76:01 And for that, we'll be using OpenAI. Let's add an Open AI. Create a chat
76:05 completion module in line just before we make the Vappy call. You'll need to set
76:08 up the connection with your OpenAI account and make sure get an OpenAI API
76:13 key and also have some credits to use for this. To do this, you can log into your
76:19 account at platform.opai.com login and click on API keys. Then create
76:23 a new one, making sure to copy and save it in a safe place. Then you'll click
76:27 over to billing and add some credits, making sure you have an active credit
76:30 card set up. With that connection set up, we'll select the model. Since we're
76:35 doing research, I'm choosing GPT40 mini search preview. By the time you watch
76:38 this, there may be other options. The important thing is to choose a model
76:42 that can do search. In the messages section, we'll set the role to user,
76:46 which just means this message is coming from a user, you in this case. The text
76:50 content is where we place the prompt. Essentially, we're telling it to serve
76:53 as a research assistant that uses our leads information to generate a summary
76:58 of how an AI automation agency could help them with their needs. Let's see
77:01 how it works by running this module only giving it an example company name and
77:11 As you can see, it's running and performing research for us. But notice
77:14 how the result is formatted in paragraphs. While that might look fine,
77:18 it actually causes a problem because when we try passing this result into our
77:22 HTTP module, it needs to follow strict formatting rules. Those paragraph breaks
77:26 can quietly break things behind the scenes and cause the system to reject it
77:30 since it won't be valid JSON. There are a couple ways to fix this. We could
77:34 either demand that chat GPT gives us our summary in JSON, or we could add a text
77:39 passer in between the GPT and HTTP modules, which removes those line breaks
77:42 for us. Since there may be other situations when you need to transform
77:46 data in your future workflows, it's helpful to get some practice passing
77:50 text. So, let's add the text passer replace tool. For the pattern, we'll add
77:53 this regular expression. This essentially finds any line breaks, so we
77:57 can replace them with a new value, which we'll leave blank. In other words,
78:01 instead of the line break, it will be replaced by nothing, aka removed. We
78:05 want global match to be set to yes, so that it finds all line breaks, not just
78:08 the first one, then no, for the rest of the options. Finally, we'll select the
78:12 text we need to pass the result from open AI's research. With this
78:16 reformatting in place, we can now reference that clean text inside the
78:22 company research field of our HTTP body. Finally, we'll say yes to pass the
78:25 response and close out this module. Notice how we still have this red error
78:31 icon on the second HTTP module. It's broken because it's trying to reference
78:35 the call ID from the original VP module that no longer exists because we deleted
78:40 that module. Instead, it needs to reference the call ID from our new HTTP
78:44 module, which manually makes that VP call. Since we haven't yet run that new
78:48 VP replacement module, we'll run it in isolation, feeding it a phone number and
78:53 whatever info just to get it to run. Once it runs, we can now access and
78:57 reference the ID field from it for future runroughs of this workflow. We're
79:02 now ready for the big payoff. Make sure you add a new lead to your Air Table
79:06 with a valid company it can research. Let's run the scenario to make sure
79:15 Hey Mark, it's Ben from Edjai. Is now a good time to talk about your business
79:19 needs? Yeah, sure is. Awesome. I'm following up on the form you submitted. We build custom AI agents
79:26 and automations for companies like View Mastery. Anything from customerf facing
79:30 bots to uh internal workflow optimization. I'd love to understand a
79:34 bit more about what led you to reach out. What sort of processes or tasks are
79:38 you looking to automate? I'm looking to implement AI agents to facilitate the
79:42 learning journey on my e-learning platform. That sounds like a great
79:46 initiative. Enhancing the learning journey for your students can really
79:49 make a difference. Are there any specific bottlenecks or manual tasks
79:54 you're hoping to reduce with these AI agents? Just hoping to increase the
79:57 effectiveness of our content by supplementing it with these AI agents.
80:03 Got it. So, you're looking to enhance the effectiveness of your content by
80:08 providing real-time support to to your students. That's a smart approach,
80:11 especially with the extensive library you have at View Mastery. Thanks for
80:15 sharing that. Would it be helpful if I sent a quick proposal your way showing
80:20 how we could support your goals? Yeah, that would be perfect.
80:23 Perfect. I'll send that to the email we have on file. Thanks again, Mark. All
80:28 right. Thanks. Exciting. Our scenario is working and our voice assistant is now
80:32 empowered to perform dynamic pitches based on our lead's unique information.
80:35 We're almost there. Finally, within build 3, we take this about as far as we
80:38 can within a beginner tutorial like this. Um, and this is a really, really
80:41 powerful thing once you've added in these extra features. So, after the call
80:44 again, we're going to analyze that and deal with if they answered or didn't
80:47 answer. And long story short, if they answered and they said that they were
80:50 interested and said, "Hey, yes, can you please send me a proposal?" Then we're
80:53 going to take all of this research that we've done and all the information that
80:56 we got from the phone call in order to generate them a custom proposal and
81:00 saying hey look this is what we want to kick off with you because you need to
81:02 make a proposal in order to start any kind of services or most kinds of
81:05 services but we can automate the generation of a proposal which can take
81:09 hours and hours and hours for businesses and we can use all of this information
81:12 we collected and they've said yes hey can you send it over and then we use an
81:15 application called Panda do and we can create a template of a document for our
81:19 agency in this case and it's going to use AI in this case we're going to use a
81:24 chatbt uh node here on make and it's going to take all of this information
81:27 and write a personalized proposal on how we would kick things off with them of
81:30 what we're proposing in terms of the scope of work for them like we will do
81:34 this this it sounds like you need this we can do this this is roughly how much
81:37 it's going to cost etc and then using panda do we can send that as a e-ign
81:40 link so that they're ready to sign and we get notifications about if they've
81:43 viewed it if they've signed it etc and so by the time you've done this we've
81:47 automated everything from the initial point of contact where the leader said
81:50 that they're interested in our services to learning more about them to
81:53 determining if they're qualified for our offer uh to sending them a custom
81:57 proposal and ultimately for them signing the dock through Panda do and they're
82:00 ready to kick things off with us. So that is an explanation of of build 3 and
82:04 what we're really trying to go here. I hope this been helpful to clarify things
82:06 for you um because this is really really powerful if you can wrap your head
1:22:09 Build #3
82:14 Once a lead has expressed interest, it's the perfect moment to harness that
82:17 momentum and transform it into something concrete, a tailored business proposal.
82:21 Why wait for someone on your team to do this manually when we already have all
82:25 the context we need? With the help of OpenAI and Panda do, we can generate,
82:30 send, and log a custom proposal without anyone lifting a finger. So, in the
82:34 final section of this course, we'll be tacking on a proposal generator to the
82:38 end of our workflow. We'll use OpenAI to create the custom text to plug into a
82:42 proposal generator. We're going to be using Panda Do as our tool for creating and sending proposals.
82:48 It allows you to create templates with placeholders that can be filled in
82:51 dynamically from your automation workflows. Here's how to create and
82:54 configure your template in Panda Do. Log into the Panda dashboard. Just create an
82:59 account if you don't yet have one. Then go to templates and click plus template.
83:03 For sake of ease, we can select an existing template to remix such as one
83:07 of these business proposal or advertising sales proposal templates.
83:10 I'm going to use a template I already created here. Within a Panda Do
83:14 template, you are able to drop in tokens, which are basically placeholders
83:18 that can be replaced with actual values, such as your client's company name. In
83:22 this case, you can set up this template however you'd like, but as you can see,
83:26 I've set mine up like this with a client introduction section addressing them by
83:30 name. In the goals and plan section, I've left room to insert information
83:34 about my lead's goals and the services I'll recommend and a plan for how I'll
83:38 implement things. so you can see how it works. We'll create those as variables.
83:43 Over here in the sidebar, we'll add proposal.goals will be a paragraph or two summarizing
83:49 the client's top priorities. Proposal.services will be a bulleted list of the
83:55 recommended services. Proposal.implementation will be a concise execution plan to
84:01 deliver the above. In the pricing section, I've already added placeholders
84:05 here. proposal pricing and for a breakdown of services and costs and
84:08 proposal.total, the full estimated project total. Down in the agreement section, we're
84:15 requesting signature and including the leads info. Once everything is in place,
84:19 name it something relevant and save it because you'll soon be using this
84:23 template inside your makes panda dooc module. Optionally, you could spend some
84:27 time styling this template with a logo, brand colors, etc. Remember, this
84:31 document will be client-f facing, so make it look and feel as professional as
84:34 the service you're offering. With your Panda do template ready and tokenized
84:38 correctly with placeholders, we're ready to include it in the final sequence of
84:42 our make scenario. We'll add the Panda Do create a document module, set up a
84:46 connection with our Panda Do account, name the document based on the company
84:49 we're sending it to, and select the proposal template we created earlier.
84:53 We'll fill in these values, giving the module the lead's email to send this to
84:59 and include all of the necessary info about our client, like the company name
85:02 and the client's first and last name. For all these proposal tokens, we'll be
85:06 generating these values with AI in a moment. For now, we'll scroll down and say yes
85:12 to send a document because we want to email this Panda doc to our lead. Fill
85:15 in the subject line using their name and company, write a short message,
85:22 then hit save. Next, we'll use OpenAI, create a chat completion module to help
85:26 us write a clear, convincing proposal. We'll select a quick and efficient GPT
85:30 model, and in the prompt, we feed it instructions about its role as a sales
85:35 expert with context about the services our mock company offers like AI
85:39 automation and agent-based systems, information about the client, including
85:43 the company research we did earlier in the workflow, and the summary of the
85:46 call our voice assistant had with them. Then we clarify its task to identify the
85:50 most relevant services we can offer them and we demand the output to be in JSON
85:54 format so we can make easy use of it in the modules after this.
85:59 If we go ahead and run this module only passing in some dummy data for the call
86:10 We'll see it efficiently generates this relevant info organized as JSON like we
86:14 requested. We're almost ready to plug these values into the Panda doc module,
86:18 but we first need to add a pass JSON module to prep the message content
86:22 output from OpenAI, breaking it up into discrete values that are available as
86:25 variables we can inject into our proposal document in the next step. To
86:29 plug these proposal values into their slots in the Panda do module, we first
86:34 need the JSON module to run. While we could run our entire scenario, there's a
86:38 quicker way. go into our OpenAI module, grab the expected JSON format, and paste
86:44 it as dummy data into the JSON module to test it. Perfect. Now, we can see it's
86:48 passing correctly, and we can now go into our Panda Do module and plug those
86:52 variables into their corresponding token Finally, we want to bring things full
87:00 circle by updating the leads air table record so that our sales team can see
87:11 So in the proposal sent on field, we'll add a timestamp with now. This keeps
87:15 everything tidy. So when someone on your team looks at a lead, they know a lead
87:19 was created, they were qualified, they were called, including a call summary,
87:22 they were interested, a proposal was sent. No one has to chase a status
87:26 update. It's all there in one glance automatically. Let's do a final run of
87:30 this scenario to make sure all of the moving pieces are working correctly.
87:37 Hey Brad, it's Ben from Edge AI. Is now a good time to quickly talk about your
87:41 business needs? Yeah, it is. Great. Thanks. I came across some info about
87:45 View Mastery. It sounds like you're really focused on helping developers
87:49 learn VDJs efficiently. We build custom AI agents and automations for companies
87:53 like like yours. Anything from customerf facing bots to internal workflow
87:57 optimization. I'd love to understand a bit more about what led you to reach
88:01 out. Are there specific processes or tasks you're looking to automate? Yeah,
88:05 I'm looking to implement AI agents for learning who are trained on the content
88:10 library of my learning platform. Uh, that makes a lot of sense. Creating
88:14 AI agents that can assist learners by leveraging your content library sounds
88:18 like a great way to enhance engagement. Are there particular bottlenecks or
88:21 manual tasks you're hoping to reduce with these AI agents? Yeah, just hoping
88:25 for uh the students to be able to comprehend and apply the concepts more
88:30 effectively with the agents. That's a smart approach. Empowering
88:37 students to better understand and apply concepts with AI support can really
88:40 boost their learning experience. Thanks for sharing that. Would it be helpful if
88:44 I sent a quick proposal your way showing how we could support new mastery with
88:49 tailored AI solutions? Yes, please go ahead and do that. Perfect.
88:54 I'll send the proposal to the email we have on file. Thanks again, Brad. Thank
89:00 you. You're welcome, Brad. Thanks for your time today. If you have any
89:03 questions later or want to discuss further, just reach out. Have a great
89:08 day. The workflow seemed to have worked as planned. Inside Air Table, we can see
89:13 that the proposal sent on date was added. Now, let's check the email that
89:16 we told Panda do to send the proposal to. Heading into that email, we can see
89:20 that it worked perfectly. Our proposal dock arrived which the client can open
89:24 up to go view their proposal which we can verify has been properly drafted
89:28 detailing the client's specific needs and our plans to help them. From here
89:32 the client can sign the document, date it and finish it out.
89:36 An added benefit of Pandanda is that we don't need to create additional
89:40 automation steps in make.com to track when leads view or sign proposals. Panda
89:44 handles this automatically by sending notification emails to our email
89:48 whenever a lead views or completes a proposal. With this final stretch,
89:52 you've created a full system that not only identifies and qualifies leads, it
89:56 translates interest into action. Of course, there are many other steps you
89:59 could add to this scenario. For example, the delay we added after the call takes
90:03 place works in most cases, but if the call exceeds 5 minutes, the system would
90:08 incorrectly mark it as not answered, even if it was successful. The foolproof
90:12 solution would be to implement a web hook that listens for the call to end,
90:16 but that's outside the scope of this beginner build. You could even add a
90:19 whole extension to this workflow where you detect when a lead signs the
90:23 proposal and then onboard them with an automated client orientation workflow. I
90:27 encourage you to get creative and add on to it to learn and challenge yourself.
90:30 For now, I want to share some troubleshooting tips to keep in mind as
90:33 you go off on your own and build your own workflows. [Music]
1:30:36 Troubleshooting & Tips
90:39 As you take your next steps and start building your own AI powered
90:43 automations, I want to be completely transparent. You will encounter issues.
90:48 platforms will change, tutorials may become outdated, and unexpected problems
90:52 arise. This isn't a flaw in the system. It's a natural part of working in a
90:56 rapidly evolving field. The truth is, even experienced developers spend a
91:00 significant portion of their time troubleshooting. I can't count how many
91:03 times I found myself yelling at my screen because some seemingly simple
91:07 thing wouldn't work. It's part of the process for everyone. There's an old
91:11 saying, give a man a fish and you feed him for a day. Teach a man to fish and
91:17 you feed him for a lifetime. This sums up our approach to technical education.
91:21 If you are spoonfed every solution, you'll end up just copying existing
91:25 systems and won't be prepared for real world challenges and you won't develop
91:28 ways to differentiate yourself either. So, think of technical problem solving
91:32 as a muscle. Right now, it might be underdeveloped and using it feels
91:36 uncomfortable. That's normal. But with consistent exercise, tackling problems,
91:40 finding solutions, learning from mistakes, this muscle will grow
91:44 stronger. It's a gradual process that develops through experience. Each
91:48 problem you solve builds your expertise, making future challenges less daunting.
91:52 While you can't rush this journey, you can fully embrace it. When you feel that
91:56 sense of being out of your depth, try to recognize it not as failure, but as
91:59 growth. You're in uncharted waters. You're pushing your limits and expanding
92:03 your capabilities. The most valuable learning happens precisely in these
92:07 moments of struggle. So with that growth mindset, let's cover some tools you can
92:12 use for troubleshooting. You probably won't be surprised that I encourage you
92:15 to leverage AI as you develop your skills. If you run into errors in your
92:19 workflow by heading over to chat GPT or a similar LLM, you can lean on its
92:23 ability to support you in solving the issues you encounter. The trick here is
92:27 to describe your issue in detail. Remember to give enough context so the
92:30 AI can grasp what you're building, the specific step you're on, and the
92:34 expected outcome versus the problem you're experiencing. Include
92:38 screenshots, error codes, and let it There will be times your issues are a
92:45 little more complicated. For complex problems, you can even try the deep
92:48 research mode to get comprehensive information on specific platforms or
92:52 issues. Sometimes the first suggested solution won't be the right one. But as
92:56 you engage in back and forth dialogue, as you see me doing here, where I am
93:03 the AI gains more clarity about your issue and can better narrow down
93:11 And if you truly want to learn, don't just do what it says without curiosity.
93:14 Ask it to help you understand why the issue happened and why the new solution
93:18 ultimately worked. If you are hitting a wall, it's possible that others have
93:22 already hit that same wall and climbed over it. So, by doing a traditional
93:25 Google search, you can look for existing threads or community posts about similar
93:30 issues. Many problems you'll encounter have already been solved by others. Of
93:33 course, YouTube is another great resource. Search for recent tutorials
93:37 that show updated interfaces and workflows or that highlight key details
93:42 that could help resolve your issue. We now have access to tools like Google AI
93:46 Studio that can watch your screen and provide real-time support as you work.
93:50 Think of this as an experienced guide peering over your shoulder and helping
93:54 you step by step. This eliminates the need to manually type out your problem
93:58 and provide screenshots to chat GPT since you can give the tool permission
94:01 to observe what's on your screen and guides you through solutions using voice
94:05 assistance. Take a look at how powerful this is as it helps diagnose what's
94:09 causing this error. I'm getting an error on this HTTP module where I'm trying to
94:13 make a post request out to the VP API. Any idea where I should start looking
94:18 for a solution to this error? I see the workflow you've created and that you are
94:21 receiving an error when trying to make a post request to the VAPI API. Are there
94:28 any error messages that you can share? It's a 400 bad request error. Do you see
94:34 the information here in red? Okay, I see that it's a 400 bad request error. The
94:39 message indicates there's an issue with the JSON formatting of your request
94:42 body. It seems like there's an unexpected character, either an expected
94:47 or after array element in JSON at position 1167, line 12, column 51. Would
94:53 you like me to help you review the JSON structure in your request to identify
94:58 any syntax errors or missing commas or brackets? Yes. Can you notice anything in here
95:03 that is wrong? Looking at the JSON, it seems like the error might be in the company research
95:09 field where you have an array containing a single element and that element seems
95:13 to be an expression. If you're trying to pass a single string there, try removing
95:17 the curly bracket. Also, can you double check that all your keys and values are
95:21 properly formatted as strings? Learning to read and apply documentation is a
95:26 critical skill. Each platform has detailed guides that are regularly
95:29 updated. Often, you'll find information there that can help solve your issue. If
95:33 it feels overly technical, you can circle back to asking something like
95:36 chat GPT to help you understand the documentation and how it applies to your
95:41 specific issue. Finally, there are plenty of online communities like the
95:45 ones me and my team run related to the platforms you're using. Join Discord
95:48 servers or other forums, search for similar issues, or even ask questions
95:52 directly. You'll be surprised by how helpful and supportive these communities
95:56 can be. If your goal is to build real automations for real clients, things
95:59 won't always go according to plan. platforms will change and requirements
96:03 will shift. The ability to adapt, troubleshoot, and find solutions is what
96:07 separates successful builders from those who give up early. The most successful
96:11 people in this field aren't those who never encounter problems. They're the
96:14 ones who persist through difficulties, who see obstacles as puzzles to be
96:18 solved rather than roadblocks to progress. So, when you hit that
96:21 inevitable moment where something isn't working, and you feel stuck, remember
96:26 this is normal. Take a breath. Step away if needed. Then come back and work
96:29 through your troubleshooting toolkit. Each time you solve a problem, you're
96:33 not just fixing that specific issue. You're becoming better at solving all
96:37 future problems. This resilience and problem solving ability will be just as
96:41 valuable to your success as any technical skill you learn. So, embrace
96:44 the journey of building your technical skills. In the next and final section of
96:49 this course, I'll show you how to sell your newly built systems to real world
1:36:55 How to Sell AI Automations
23:45 What We're Building
23:49 before we jump in, it's very important that you understand what we're doing in
23:52 this first build, but also in the second and the third and how it all fits
23:55 together and you get kind of an idea of where we're going. Understanding this
23:57 technology and then being thrown in a whole bunch of random things uh like Air
24:01 Table and and Slack and make and all of these different terms can be a bit
24:04 confusing. So, I just want very quickly to give you a bit of an orientation
24:07 before we jump into things of what we're building and why and how it works
24:09 because I've touched on it before, but I don't think I did a good enough job. So,
24:12 I'm filming this after just to really make sure you guys are fully clear on
24:15 what we're building here. So, and why it's valuable really as well. So um as
24:20 we touched on this is going to be an AI qualification system um for inbound
24:24 leads to a business. So easiest way to explain this is if we had a a website
24:28 and say we had this book a free consultation form and we might be
24:31 running ads to it. We might just have that on a website and people just
24:34 discover it by searching the web or or we're showing up in Google results. But
24:37 at some point, a business like the example I gave before, if you're an
24:41 accountant and you only work with uh builders or plumbers or whatever the the
24:44 example I gave, some people are going to fill out this form who are not
24:47 necessarily the right person for you. Like the my accountancy firm that only
24:51 works with plumbers, if a surgeon comes in or if a a pilot comes in and fills
24:54 this thing out and wants to get accounty services, we would ask them questions
24:57 around, okay, well, what's your name? What what do you do? Uh what sort of
25:00 industry do you work in? And we'll collect this information here. And as I
25:04 mentioned before, not all of the people who fill out your forms, particular if
25:07 you're running ads or or doing content and it's driving traffic to this landing
25:10 page as a business, not all of them are going to be people that you actually
25:13 want to get on calls with because if you get on calls with them and they're not
25:16 your ICP or the person that you could potentially work with, then it's going
25:19 to be a complete waste of time. So, a qualification is a very essential part
25:22 in in pretty much any business. And what you're going to build is a very very
25:24 powerful system starting off with something basic like this. And then
25:27 we're going to progress it to something a bit more advanced like this. And then
25:30 finally, something extremely valuable and powerful that you guys are all going
25:33 to be able to build along and ultimately sell this to someone if you wanted to
25:35 afterwards. So that's going to be covered in the monetization section. But
25:39 just to clarify sort of what we're going to be doing here in this first build, uh
25:43 you'll start off with building a form on a telly. So we're essentially
25:45 replicating this kind of website form, but we're just going to build it in
25:48 tally in this case. You can embed these into different websites, etc. But just
25:51 think of this as a form that people will get sent to. So maybe they clicked on an
25:54 ad and then they get sent to this. How can we help? We're going to build out
25:57 this form and then when people click submit on this form, it's going to send
26:00 them into this basically a fancy spreadsheet. And in this case, we're
26:03 using Air Table. Air Table is like a really awesome platform which you're
26:05 going to get to to use a lot within this build. Uh but Air Table, you can think
26:09 of it like a fancy spreadsheet and we're going to be able to map uh each of these
26:13 answers to a row in this database here. So every time someone fills this form
26:16 out that we create, it's going to load it and add it as a row into this
26:20 database. And the cool thing there is that air tableable allows us to within
26:24 make.com our workflow builder it allows us to set up events that when things
26:27 happen in air table like okay let's watch for new records that arrive and
26:31 that's what this here is doing. So this trigger of our automation is going to be
26:35 watching for new records that arrive in this and then it's going to take
26:39 anything that arrives in here say this first row and then it's going to put it
26:42 through this automation here. And so this is going to be a very basic one
26:45 where we've taken actually the AI component out of this. Before Air Table
26:48 really got advanced with the AI features. What you're doing here is take
26:51 the information from the Air Table. You'd use some chat GPT step here and
26:54 then you'd be able to do things afterwards. But in this case, I'm making
26:57 it extra simple for you guys and that we're actually just using Air Table's
27:00 built-in AI features. So down here, you can see in the tutorial that you're
27:03 about to go into, we will use Air Table to create an AI column here that's going
27:07 to take in information like uh like here where they're talking about their their
27:10 company, their budget, and about their needs. We can write a prompt in here and
27:13 it's going to automatically fill out this row. So as soon as a new person
27:17 arrives in this spreadsheet, this field that we're creating here is going to
27:20 automatically qualify them and use AI and this prompt that we write here to
27:23 analyze the information that they gave us and then output if they are qualified
27:27 or if they're not. So we're doing the qualification with AI built into Air
27:30 Table which is going to save us a lot of time and messing around with jumping
27:33 back and forth. So So that is basically the start of build one. We will do
27:37 immediate qualification with AI within Air Table and then we're going to if
27:40 they are qualified we're going to send an email to them and because we're
27:43 already doing the qualification here in Air Table it makes it a lot easier for
27:46 us over at make.com because for this particular node as you'll see in a
27:49 second we're going to set it up so that it's only actually going to look for uh
27:54 the rows or records in this table that have been AI qualified as qualified. So
27:58 if this prompt here that we've written analyzes all the information in the row
28:02 and then outputs qualified then and only then will it pass it into here. And this
28:05 is when we're going to be sending an email to the prospect and saying, "Hey,
28:09 we're interested in in hopping on a call. Here's a link to book in a call
28:12 with us." But at the same time, it'll also send a message to our sales team in
28:15 Slack and say, "Hey, look, we've just had a new qualified lead." So, this is
28:19 very much a a minimized version of what you could do uh maybe a year ago. But
28:24 because of the Air Tables AI features, uh we can do a lot of the heavy lifting
28:27 here and writing a prompt that's going to analyze what the person has filled
28:30 out here, aka are you a are you a builder or are you a pilot? And if they
28:34 were a builder, in the case of the example I just gave, it's we're going to
28:36 be using a different example in the build. Um, but it would come out as
28:39 saying qualified here. And because it says qualified, then it would
28:43 automatically trigger uh the rest of this automation, which is sending an
28:45 email to them and taking the next step and saying, "Yes, we are interested in
28:48 talking to you further, making sure that we're not talking to anyone that we
28:51 shouldn't be talking to and at the same time letting the sales team know." So,
28:54 that is build one. Um, that is how it all fits together. And I'm going to be
28:57 doing these little updates just before each build so that you guys are 100%
2:26 Why Learn AI Automation?
2:26 of this. So let's dive in. So there's a lot to cover here. I don't want you to
2:29 give up halfway. So let's quickly get clear on why learning AI automation is
2:33 one of the most valuable skills anyone can have over the coming decade. Whether
2:36 you're a student, an employee, or an entrepreneur. Here's some quick truths
2:40 about AI and jobs. McKenzie predicts that AI and automation can replace up to
2:45 50% of current work activities by 2030. And the World Economic Forum states that
2:49 41% of companies plan to reduce staff due to AI. Now, this is a lot of doom
2:52 and gloom and many are naturally worried about their career in the future when
2:55 they hear the stuff, but it's not actually all bad if you know where to
2:58 look. So, on the flip side of this same data, these same reports reveal an
3:02 enormous opportunity for those willing to seize it. The board economic forum's
3:06 future of job report states that 50% of employees plan to reorient their
3:09 business in response to artificial intelligence and 66% of employees plan
3:15 to hire talent with specific AI skills such as AI workflow automation. So on
3:18 one hand we have the expectation of massive layoffs and automation of work
3:22 over the next 5 to 10 years. But on the other we have the majority of employers
3:26 searching for people who have AI skills or really just some form of basic AI
3:29 literacy. Why is this? Well, it's because AI literate individuals who can
3:33 identify opportunities for automation and automate them themselves can have 5
3:36 to 10x the output of someone who doesn't know this and can't automate their own
3:40 work. And I promise you that brushing up on your AI and actually becoming AI
3:43 literate so that you can be on the winning side of this next 5 to 10 years
3:47 is so much easier than you think. I mean, it's it's literally as easy as
3:51 watching this entire video in order to build your AI skills base. If you don't
3:54 believe me when I say that a little bit of self-study like this video goes a
3:57 long way, here is an excellent clip from the All-In podcast from one of the most
4:01 respected investors and technologists in the world, Naval Ravakant, alongside a
4:05 whole bunch of other biners. Again, I would say the easiest way to see that AI
4:08 is not taking jobs or creating opportunities is go brush up on your AI,
4:14 learn a little bit, watch a few videos, use the AI, tinker with it, and then go
4:18 reapply for that job that rejected you and watch how they pull you in. This
4:21 video is exactly what Naval is talking about. This is why I create these. So
4:24 whether you're a student wanting to stand out in a competitive job market or
4:28 an employee aiming to become irreplaceable at work or an entrepreneur
4:31 like me looking to scale your business with cuttingedge tools and automations,
4:35 I have made this video for you. Now close out of all your other tabs, get a
4:38 notebook and a pen and a beverage of your choice and make a commitment right
4:41 now to yourself to finish this training and to ensure that you're going to be
4:45 empowered by AI and not replaced by it. That is all I want out of this video for
4:47 you all. So if you've done all that, let's get stuck into it.
4:49 What is AI Automation?
4:53 All right. So, step one in building AI automations is knowing what an
4:56 automation actually is. And the term gets thrown around a lot these days.
4:59 First thing we need to realize is that the AI part of the term is is relatively
5:03 new. Automation itself has been around for a long time. So, let's start there
5:07 to make this super easy to grasp. In simple terms, an automation is a system
5:10 that does a task for you without you having to lift a finger. It's kind of
5:13 like setting up a little robot to do boring, repetitive stuff automatically,
5:16 so you don't have to waste your time on it. These are what we'll call old school
5:20 automation. The kind that existed way before chat GBT came along. They were
5:23 often built on platforms like Zappia and lots of small to medium-sized businesses
5:27 used them for the past 5 10 years. And they did basic things like automatically
5:31 saving info, for example, when someone filled out a form on a website. You
5:34 could make an automation that would take their name and their email and then just
5:37 pop it into a spreadsheet. So just a little automation automating that boring
5:40 stuff. Or for example, when an email came in, it would send a quick alert to
5:44 a chat app like Slack. So it's kind of like having a little helper who's
5:46 following a super simple checklist. If this happens, then do that. No thinking,
5:50 just just doing, right? And the benefit of this is huge. It freed humans from
5:54 doing super basic and boring work. And for business owners, it meant not having
5:57 to pay more people just to handle these tiny and annoying tasks. It's kind of
6:01 like having a tireless assistant who never complains about doing the same
6:04 boring thing over and over and over again. So for decades, all was well in
6:08 the automation space. And these old school automation saved time and money,
6:11 and everyone was super happy. That was until the release of Chat GBT in late
6:16 2022. It blew the entire field wide open, turning automation from some niche
6:20 trick used by some savvy companies into the biggest thing since the internet.
6:24 Generative AI models like chat GBT added to automation was like putting a V12
6:27 onto a bicycle for this automation space. They made it possible to do more
6:31 than just simple tasks. This was because these powerful AI models could handle
6:34 much trickier stuff that used to require a human brain. Instead of just updating
6:38 a spreadsheet row automatically, these automation platforms can now use the
6:42 power of Chat GBT to do things that only people could have done before. So
6:45 platforms like make.com, which you'll be learning more about later in this video,
6:48 enable us to automate things like write a whole post for LinkedIn sounding just
6:52 like you. Pulling out names, places, and phone numbers from giant documents in
6:55 seconds. Reading and figuring out if an incoming email is someone asking for a
6:59 refund or just wondering where their order is. Shrinking huge piles of info
7:03 into short and easy to read reports, spotting things in the picture like
7:06 identifying a product in a photo, or even creating brand new images and
7:10 videos from just a few words. So what chat GPT and the explosion of other
7:14 amazing generative AI tools gave us was basically human intelligence on demand.
7:17 These AI models are kind of like having a super smart friend who can do almost
7:21 anything that you ask them as long as you tell them clearly what you want. And
7:24 using automation platforms, we can easily set up these super smart friends
7:28 into our systems and use these kinds of models in thousands of different ways.
7:32 All you have to do is pick the right AI tool for the job, give it a clear
7:35 instruction, call a prompt, and watch the magic happen before our eyes. And
7:38 that is how the AI automation industry was born. So, with that little history
7:42 lesson out of the way, let's get back to our original question of what is an AI
7:46 automation. Well, it turns out that this field is so new that there isn't even an
7:49 official definition for what an AI automation is. So, here's mine. Just
7:53 keeping it nice and simple. An AI automation is a system that uses AI to
7:57 automatically do complex tasks that would normally require a human. So, the
8:00 big difference between these old school automations that we've just talked about
8:04 and today's AI automation is the kinds of tasks that they can handle. Thanks to
8:08 these recent advances in AI technology, we've gone from just moving data around
8:11 and putting stuff in spreadsheets to being able to solve problems that need
8:14 thinking and creativity and decision-making. It's kind of like
8:17 upgrading from a basic toy robot that only moves forward to a high-tech robot
8:21 that can solve puzzles and move around the world. So, when you learn AI
8:24 automation, you are basically learning how to build digital workers that can do
8:27 very powerful things for you without ever having to lift a finger. This is
8:31 why so many people are racing to pick up the skill right now before it's too late
8:34 to stay ahead. It's like the ultimate cheat code because you can build them to
8:38 do exactly what you need, tailored to any kind of job or workflow. For
8:41 example, for students, it's kind of like having a helper that can organize your
8:44 study notes automatically. Or for employees, your AI automations can be
8:47 like a buddy that handles the boring paperwork so you can shine on bigger
8:50 projects. And for entrepreneurs, it's like having a team member who runs your
8:53 business tasks while you sleep. And the great thing is that they all cost way
8:56 less than hiring extra people. And they don't need breaks or vacation days, and
8:59 they don't grumble about doing the boring stuff. So, I'm sure you can see
9:03 why businesses and students and anyone looking to save time and money are so
9:07 excited about this technology and why anyone who knows how to build these
9:11 systems is instantly 10x more valuable. Just imagine automating something simple
9:15 in your life or work like sorting emails or scheduling tasks and how much time
9:19 you'd be able to save to focus on what really matters in your life.
9:24 Now, before we dive deeper, it's important to understand that AI
9:27 automation is a super broad term these days, covering wide ranges of different
9:31 systems and applications that can be built with AI. This is largely due to
9:34 the rapid advances in areas like AI agents and AI tools. So, over the past 2
9:39 years, as I built my own AI agency, Morningside AI, and helped thousands
9:42 through my communities to do the same, I've had to create a clear system for
9:46 making sense of the chaos that is the AI automation landscape. And I really want
9:49 to share that with you today because it's been super helpful for me and many
9:52 of my students to be able to figure this space out and at least have some mental
9:55 buckets that you can put things in. Here's the three different categories
9:57 that you need to keep in mind. And please stick with me. This will all make
10:00 sense in a second. So firstly, we have conversational AI. These are systems
10:03 that chat with people handing back and forth conversations. It's kind of like
10:07 having a friendly robot that talks to customers for you. These kinds of chat
10:09 bots can be found on things like websites and answer questions or they
10:13 can be voice agents that pick up phone calls. These used to need real people to
10:16 talk, but now AI can automate these kinds of conversations. The second
10:19 category is AI tools, and these are systems that use AI to do a specific job
10:23 when a person asks them to, and it's mostly to help workers get more done.
10:27 For example, I can make a custom AI tool that takes a link to a cool blog post
10:31 that I found, grabs the info from the web page, does extra searches on the
10:34 topic, and then uses something like chatbt to write a new beta version for
10:39 my own blog. And third and final is AI workflow automations. These are systems
10:43 that do a whole series of tasks by themselves, starting when something
10:46 happens, like a trigger or on a set schedule like once a day. They use AI to
10:50 make decisions that used to need a human brain. It's kind of like having a smart
10:53 robot manager that runs the whole process for you. For example, an
10:56 automation can call customers of an online store 14 days after they buy
11:01 something using an AI voice agent to ask for feedback and review all without you
11:04 having to do a thing. So, you set it up on that trigger of 14 days after
11:08 purchase, then execute this workflow. Now, those three categories might be a
11:10 little bit confusing if you are completely new to the space, but don't
11:13 worry. In the building section of this video, we're going to be creating three
11:16 automations which integrate each of these different types, and you get to
11:19 see them in action, which will make it super clear. So, AI automation is
11:22 essentially an umbrella term under which all of the exciting stuff in the AI
11:25 space is happening right now. However, when people talk about AI automations,
11:28 they are typically referring to the last type that I mentioned. So, AI workflow
11:31 automation, and these are what we're going to be building later in the video.
11:35 And an automation refers to one chain of steps that uses AI in various ways to do
11:39 certain tasks. And if I'm honest, this last category of AI workflow automation
11:42 is really the most powerful because it can incorporate all elements of AI
11:46 automation like agents, conversational AI, and tools in order to build
11:50 end-to-end processes that are much more valuable than if they were just alone.
11:52 So, what I'm saying is that what you're about to learn is the key skill that
11:55 lets you do pretty much anything in the AI space these days. It is the
11:59 foundation for building systems that save time, make money, or just make life
12:03 easier, whether you're at school or work or running your own show. And after
12:05 years in the game, I can tell you that it's been one of the most valuable
12:10 skills that I have ever picked up. So, now that you understand what AI
12:13 automations are, let's take a little peek under the hood and see how they
12:16 actually work. So, don't worry if this sounds tricky. I've been breaking down
12:19 this kind of complex AI stuff for years now. So, I'm going to make this super
12:21 easy to understand for you. So, you can think of an AI automation like a facto's
12:25 assembly line, right? There's different stations and they're all working
12:28 together to build something awesome from start to finish. It's kind of like
12:31 having a a team of little robots, each with a special job passing the project
12:35 along until it's done. So, let's go through the five key parts to make this
12:38 magic happen. Firstly, we have the trigger. This is the very first step of
12:41 an automation. You can think of it as the facto's start button or the whistle
12:45 that says, "Let's go." It's what kicks everything into gear. It could be
12:48 something like a new email popping into your inbox, a form being filled out on a
12:53 website, or even a specific time of day. This is more of a a schedule. Secondly,
12:57 we have a filter. So, not everything that starts in the automation should
13:00 keep going through it. So, a filter essentially checks if what came in is
13:03 the right stuff to work on. It's like how a factory worker does some kind of
13:06 quality control and checks that if the materials that they've received are good
13:10 enough to use in the final product. If they are not then they get tossed out
13:12 and if they are then they move forward to the next part of the sequence. You
13:15 can think of it kind of like a bounce at a club where only the important stuff
13:18 and the good things that you want inside the club or in your automation are
13:21 allowed through. Thirdly, we have actions. So this is where the real work
13:25 gets done. Actions are the steps your automation takes like the different
13:28 stations in a factory where each one does a specific job. So for example,
13:31 your automation might send an email, update a list or create some kind of
13:34 report. Often there are going to be several actions one after another just
13:37 like a product moving down the line getting built bit by bit in order to
13:41 achieve one of these outcomes. Next we have the intelligence layer. So this is
13:44 where the AI magic shines. This part is like having a super smart robot on the
13:48 assembly line that can think, analyze, and make decisions on the spot. And you
13:51 can tell it how to think using prompting. The AI inside your automation
13:55 can look at each task, figure out what's needed and adapt based on the context
13:58 you provided, like deciding how urgent something is, writing a custom message,
14:02 or pulling out key info from a big mess of data. These intelligent steps can go
14:06 way beyond just following some kind of preset rules that we saw with old school
14:09 automations. Next, we have the format. So, just like items in production line
14:13 may need some sanding or trimming before another piece can be added on top, the
14:16 data in our automations often need some kind of adjustments along the way. And
14:19 this is where we use a formatter to prepare things for the next step. And
14:22 finally, we have the output. This is your finished product, just how Carac
14:25 Factory packages up the final item and ships it out. Your automation delivers
14:28 the completed work at the end of the sequence. This could be a message sent
14:32 to your team, an updated file in your system, or a finished document that's
14:35 ready to go. At the end of the line, it's where everything comes together.
14:38 It's like after making pizza and putting all the toppings on and prepping the
14:40 dough, you're finally getting out of the oven. It's hot. It's ready to eat and
14:43 ready to go. So, let me show you how all of these parts work together with a real
14:46 example that anyone in a job can relate to. Say you're an employee who wants an
14:50 automation to handle incoming customer emails while at work. So the trigger is
14:53 going to be when a new email lands in your company's support inbox. The filter
14:56 is going to check if it's something important like if the email mentions
15:00 urgent or problem then it will pass it along to the next step. If not, it will
15:03 end the sequence there. Thirdly, we have the intelligence layer which is going to
15:06 use AI to read the email, figure out what it's about and then helps to draft
15:10 a helpful response. And in this case, the intelligence layer also acts as a
15:13 formatter which is essentially packaging the response and making sure it's in the
15:16 format we want. Then the actions in the automation can actually send that reply
15:20 and alert the boss on Slack if needed. The output logs everything neatly in
15:24 your system and marks it as handled. So that's just a simple example, but these
15:27 same building blocks are used to automate much bigger things too, which
15:31 is why AI automations are such a game changer these days. They combine these
15:34 parts to create truly smart systems that can save time and boost efficiency,
15:38 whether you are you're juggling tasks at work or crushing it in school or running
15:40 your own business. So, now that you've got a handle on what AI automations are
15:44 and how they're built, let's take a quick look at the tools that make all of
15:48 this possible. Now, don't worry if this sounds a little bit too techy. Soon, these tools are
15:53 going to literally be second nature for you, and it's going to feel so easy to
15:56 do. Creating automation starts with picking a main automation platform.
15:59 These platforms are typically called workflow builders, and they are the
16:03 command center of your automation factory. They give you a blank canvas to
16:06 design your automation on using the building blocks we talked about. Some of
16:09 which are going to be powered by AI like chatb. Popular workflow builders include
16:13 make.com which we're going to be using later in this video in the tutorial
16:16 section. We have Zapia which is great for quick setups and we have NAM which
16:19 is perfect if you want a bit more control. They are the brain of your
16:22 operation and they're really just controlling how everything fits
16:25 together. So workflow builders don't work alone. They are essentially a place
16:28 to hook up all sorts of other tools to get the job done. So these are the
16:30 categories of tools which you can connect to your workflow builders.
16:33 Firstly, we have databases and spreadsheets. So for storing information
16:36 and data, you can use things like Air Table or Google Sheets. You can think of
16:39 them as basically the filing cabinets where you keep all your data neat and
16:42 tidy so that you can save new things to the database or you can pull it into
16:45 your automations as needed. Secondly, we have communication tools for sending
16:49 messages. Things like Slack or Google. They're like kind of walkie-talkies that
16:51 can pass information around automatically for you. Then we have AI
16:55 models. And these can add that smart human level thinking like open AI's chat
16:58 GPT which you can think of kind of like a genius buddy who solves problems for
17:02 you when you give him the instructions that he needs. Then we have scheduling
17:05 tools which can handle time and meetings things like calendarly or Google
17:07 calendar and these are essentially like your personal planner to keep things
17:10 running on schedule. Then we have forms and intake too. So this can collect
17:13 information from people where you have things like type form or tally. These
17:17 are essentially input forms and triggers for your automations when someone fills
17:19 them out. So as your automation goes through it's like calling up all your
17:22 friends on a group project. Each one is bringing their own special knowledge or
17:26 capabilities to the table in order to help you to get to the end of that
17:28 sequence and finish off the automation to create the final product. By building
17:31 out a workflow, you are basically the boss at the factory. Once you know what
17:35 each tool can do, you can mix and match them to work together smoothly along
17:38 this assembly line. It's kind of like building with Lego blocks. You can just
17:42 like snap the right pieces together to make something awesome. You get to
17:45 decide which tools you connect in what order and how AI can make the whole
17:49 system smarter as well. Which AI models do you use? Gemini, do you use OpenAI?
17:52 to use complexity for searching. There's so many different decisions you have to
17:55 make. Uh but it's really cool to be able to flex your kind of creativity in order
17:58 to solve these problems. It's really like a a new age form of problem
18:01 solving, which is why I love it so much. I say, I know I need to take this and
18:05 get to this. How can I use AI to do that? And really forces you to explore
18:08 what's out there in terms of AI tools these days. So, what kinds of workflows
18:11 can we build? So for students, for example, you could build a study
18:14 material organizer that automatically summarizes lecture recordings for you,
18:18 creates flashcards from notes, and schedules review sessions based on your
18:21 exam dates. An employee could improve their workflow with a meeting assistant
18:24 that records and transcribes their meetings, and then generates action item
18:28 summaries off that and updates project management tools. Entrepreneurs could
18:31 design a lead qualification system, which automatically qualifies leads,
18:35 calls them with an AI voice agent, and then sends them a custom proposal to
18:38 automate that whole process. And that is exactly what you'll learn how to build
18:41 by following along with me in this next section. So, I'm going to show you how
18:43 to build out a lead qualification automation step by step, starting simple
18:48 and getting more complex as we go. We're going to be focusing on this
18:50 business-based workflow because these kinds of business systems offer the best
18:54 opportunity for monetizing your new skills, which I'll go into depth on at
18:57 the very end. So, stick with me and you're going to be learning how you can
18:59 start to make money immediately with these kinds of skills. So, before we get into that, let's do a
19:05 quick summary of this section. So firstly, an AI automation is a system
19:09 that uses AI to automatically perform complex tasks that used to require
19:14 humans. Secondly, we build them inside of workflow builders that incorporate
19:17 integrations with tons and tons of other tools. And every automation has six key
19:21 components. Firstly, a trigger, what starts the workflow, a filter,
19:25 conditions that need to be met, the intelligence layer, which processes info
19:29 and makes decisions with AI, the actions, which are the tasks that
19:32 actually get performed, formatterers, which clean things up, and the output.
19:36 the final result or deliverable. So, if you're feeling unclear about anything
19:39 we've covered so far, feel free to go back and rewatch some of those sections
19:42 and be ready to join us in the next step when we begin building. So, it's really,
19:45 really important that you do understand everything that we've gone over there
19:48 because you're not going to have a solid foundation to build your technical
19:51 skills on top of, which we're doing in the next section. So, please, please,
19:54 please, I've taken a lot of time to put all of this information very, very
19:57 gradually together. So, you must understand everything that I've gone
20:00 over just here before we go into the next phase. If that's all good, then
20:02 let's take a look at what we're going to Okay, so now that we have the
20:08 foundational knowledge built that you need, we can now get into the second
20:11 chapter of this video where we're going to be building three AI automations from
20:14 scratch. We're going to be starting off with something super beginner friendly
20:17 to get you started and then working our way up to a much more advanced and
20:20 valuable one by the end. Now, very important note is that each of these
20:23 automations build on each other. So, you have to be able to do the first one in
20:26 order to be able to make the second and so on. So, you you cannot skip ahead.
20:30 I've planned this out very carefully to gradually layer on your skills. So, it's
20:33 all very intentional. So, please just trust the process. So, over the next
20:36 chapter, you'll learn almost all of the key skills you need to start building
20:39 your own AI automations from scratch and be able to tap into this enormous
20:42 opportunity that is AI automation. The system that we're going to be building
20:45 gradually over the next three sections is an AI lead qualification and proposal
20:49 generation system for a business. In the first tutorial, we'll be setting up the
20:53 base using AI to automatically qualify leads after they submit a form on their
20:56 website. In the second tutorial, we'll improve that qualification ability by
21:00 implementing an AI voice agent that can call the lead for more information. And
21:03 finally, in the third tutorial, we'll implement an automated proposal
21:06 generator that can instantly create proposals for qualified leads using the
21:09 information collected on the phone call and in the lead form. The whole point of
21:12 this is that as soon as someone's interested as contacting the business,
21:15 they can kick things off with them immediately by getting a proposal in
21:17 their hand. So this process of qualification is a crucial part of
21:21 running any business at scale because at the end of the day, not all people who
21:24 come to a business are going to be a good fit for their services. So for an
21:27 example, someone may run an accountancy business, but they only choose to work
21:30 with doctors. They are specialized in helping doctors with their finances. But
21:33 if a builder fills out their website form, then that lead would not be
21:37 qualified, right? Taking a call with them would be a waste of time because
21:40 they are not a doctor. Therefore, to stop wasting time, almost all businesses
21:43 need some kind of qualification system. And for qualified leads, most businesses
21:47 need to make some kind of proposal for them in order to kind of look at it and
21:50 and see what they're proposing and then agree to that proposal. Since most
21:54 proposals don't lead to a deal, maybe 20% if you're lucky, this is a huge
21:57 waste of time and resources for the company. So, long story short, what
22:00 we're about to build will solve a number of key problems for basically any
22:03 business, making it extremely valuable for you to be learning how to do it. And
22:05 in fact, this is something you'll be able to go and sell directly to
22:08 businesses when you're done. Here's what this process looks like without AI
22:11 automation first. So, a human sales rep must constantly check for new form
22:14 submissions from the website, review each lead's details, and evaluate if
22:18 they're worth pursuing, research the company to understand how to help them,
22:21 make phone calls, and deliver competent pitches, and then manually create custom
22:25 proposals. That is hours of repetitive work per week. And it could lead to
22:27 potential leads slipping through the cracks. If maybe they get the
22:30 qualification wrong, they don't understand what the business is doing,
22:32 they don't properly research it, or they're just too slow at getting back.
22:35 They may have to wait a whole day for the lead to hear anything back. So
22:37 decreasing the time it's going to take for a qualified lead to hear back from
22:41 them. It's going to drastically increase their conversions. With AI automation,
22:44 we can transform this tedious process into something much more efficient,
22:46 which is exactly what we're going to be building together. So here's how it
22:49 works. It's going to start as usual with a lead filling out the form on the
22:52 website. Then the system is going to automatically qualify them using AI.
22:56 Then it's going to further research the company using AI also. Then it's going
23:00 to send an automated phone call to pitch your offer using an outbound AI voice
23:04 agent. The system then saves the call outcome, summarizes the conversation,
23:07 and generates a personalized proposal for them, all without a human having to
23:11 lift a single finger and in a super scalable way. This is the power of AI
23:15 automation. We're taking tasks that used to demand hours of manual work and
23:18 turning them into workflows that run all by themselves. and we're going to build
23:21 it all using several of the most popular tools from the automation ecosystem. So,
23:24 if you're as excited to learn this as I'm to teach you, then let's get
23:27 started. And everything that you need to follow along with it, including all
23:30 resources, templates, prompts, etc., is going to be available for free on my
23:33 school community. You can find it in the first link in the description. You'll
23:36 need to request to join. It will take 1 to 2 minutes to be accepted. Once you're
23:39 in, you can just search for the title of this video, and then you'll be able to
23:42 find all the resources attached to it. So, that is how you can get all the
23:44 resources to follow along. Let's get stuck into it. All righty, guys. Just to clarify things
23:49 before we jump in, it's very important that you understand what we're doing in
23:52 this first build, but also in the second and the third and how it all fits
23:55 together and you get kind of an idea of where we're going. Understanding this
23:57 technology and then being thrown in a whole bunch of random things uh like Air
24:01 Table and and Slack and make and all of these different terms can be a bit
24:04 confusing. So, I just want very quickly to give you a bit of an orientation
24:07 before we jump into things of what we're building and why and how it works
24:09 because I've touched on it before, but I don't think I did a good enough job. So,
24:12 I'm filming this after just to really make sure you guys are fully clear on
24:15 what we're building here. So, and why it's valuable really as well. So um as
24:20 we touched on this is going to be an AI qualification system um for inbound
24:24 leads to a business. So easiest way to explain this is if we had a a website
24:28 and say we had this book a free consultation form and we might be
24:31 running ads to it. We might just have that on a website and people just
24:34 discover it by searching the web or or we're showing up in Google results. But
24:37 at some point, a business like the example I gave before, if you're an
24:41 accountant and you only work with uh builders or plumbers or whatever the the
24:44 example I gave, some people are going to fill out this form who are not
24:47 necessarily the right person for you. Like the my accountancy firm that only
24:51 works with plumbers, if a surgeon comes in or if a a pilot comes in and fills
24:54 this thing out and wants to get accounty services, we would ask them questions
24:57 around, okay, well, what's your name? What what do you do? Uh what sort of
25:00 industry do you work in? And we'll collect this information here. And as I
25:04 mentioned before, not all of the people who fill out your forms, particular if
25:07 you're running ads or or doing content and it's driving traffic to this landing
25:10 page as a business, not all of them are going to be people that you actually
25:13 want to get on calls with because if you get on calls with them and they're not
25:16 your ICP or the person that you could potentially work with, then it's going
25:19 to be a complete waste of time. So, a qualification is a very essential part
25:22 in in pretty much any business. And what you're going to build is a very very
25:24 powerful system starting off with something basic like this. And then
25:27 we're going to progress it to something a bit more advanced like this. And then
25:30 finally, something extremely valuable and powerful that you guys are all going
25:33 to be able to build along and ultimately sell this to someone if you wanted to
25:35 afterwards. So that's going to be covered in the monetization section. But
25:39 just to clarify sort of what we're going to be doing here in this first build, uh
25:43 you'll start off with building a form on a telly. So we're essentially
25:45 replicating this kind of website form, but we're just going to build it in
25:48 tally in this case. You can embed these into different websites, etc. But just
25:51 think of this as a form that people will get sent to. So maybe they clicked on an
25:54 ad and then they get sent to this. How can we help? We're going to build out
25:57 this form and then when people click submit on this form, it's going to send
26:00 them into this basically a fancy spreadsheet. And in this case, we're
26:03 using Air Table. Air Table is like a really awesome platform which you're
26:05 going to get to to use a lot within this build. Uh but Air Table, you can think
26:09 of it like a fancy spreadsheet and we're going to be able to map uh each of these
26:13 answers to a row in this database here. So every time someone fills this form
26:16 out that we create, it's going to load it and add it as a row into this
26:20 database. And the cool thing there is that air tableable allows us to within
26:24 make.com our workflow builder it allows us to set up events that when things
26:27 happen in air table like okay let's watch for new records that arrive and
26:31 that's what this here is doing. So this trigger of our automation is going to be
26:35 watching for new records that arrive in this and then it's going to take
26:39 anything that arrives in here say this first row and then it's going to put it
26:42 through this automation here. And so this is going to be a very basic one
26:45 where we've taken actually the AI component out of this. Before Air Table
26:48 really got advanced with the AI features. What you're doing here is take
26:51 the information from the Air Table. You'd use some chat GPT step here and
26:54 then you'd be able to do things afterwards. But in this case, I'm making
26:57 it extra simple for you guys and that we're actually just using Air Table's
27:00 built-in AI features. So down here, you can see in the tutorial that you're
27:03 about to go into, we will use Air Table to create an AI column here that's going
27:07 to take in information like uh like here where they're talking about their their
27:10 company, their budget, and about their needs. We can write a prompt in here and
27:13 it's going to automatically fill out this row. So as soon as a new person
27:17 arrives in this spreadsheet, this field that we're creating here is going to
27:20 automatically qualify them and use AI and this prompt that we write here to
27:23 analyze the information that they gave us and then output if they are qualified
27:27 or if they're not. So we're doing the qualification with AI built into Air
27:30 Table which is going to save us a lot of time and messing around with jumping
27:33 back and forth. So So that is basically the start of build one. We will do
27:37 immediate qualification with AI within Air Table and then we're going to if
27:40 they are qualified we're going to send an email to them and because we're
27:43 already doing the qualification here in Air Table it makes it a lot easier for
27:46 us over at make.com because for this particular node as you'll see in a
27:49 second we're going to set it up so that it's only actually going to look for uh
27:54 the rows or records in this table that have been AI qualified as qualified. So
27:58 if this prompt here that we've written analyzes all the information in the row
28:02 and then outputs qualified then and only then will it pass it into here. And this
28:05 is when we're going to be sending an email to the prospect and saying, "Hey,
28:09 we're interested in in hopping on a call. Here's a link to book in a call
28:12 with us." But at the same time, it'll also send a message to our sales team in
28:15 Slack and say, "Hey, look, we've just had a new qualified lead." So, this is
28:19 very much a a minimized version of what you could do uh maybe a year ago. But
28:24 because of the Air Tables AI features, uh we can do a lot of the heavy lifting
28:27 here and writing a prompt that's going to analyze what the person has filled
28:30 out here, aka are you a are you a builder or are you a pilot? And if they
28:34 were a builder, in the case of the example I just gave, it's we're going to
28:36 be using a different example in the build. Um, but it would come out as
28:39 saying qualified here. And because it says qualified, then it would
28:43 automatically trigger uh the rest of this automation, which is sending an
28:45 email to them and taking the next step and saying, "Yes, we are interested in
28:48 talking to you further, making sure that we're not talking to anyone that we
28:51 shouldn't be talking to and at the same time letting the sales team know." So,
28:54 that is build one. Um, that is how it all fits together. And I'm going to be
28:57 doing these little updates just before each build so that you guys are 100%
29:05 To get started, we need a way to gather our leads information. While there are
29:09 several great options for this build, we'll be using Tally, an easy to use
29:13 form builder. Just sign up for free or Then we'll click on new form and give
29:20 our form a title. Anyone using the form will see this title. So, we want it to
29:24 make sense for the user. So, in our case, we can name it, how can we help? I
29:27 want to build it from scratch. So, I'll hit enter. The way we create forms in
29:31 tally is with building blocks. So, if we click the plus button, we can select the
29:35 type of block we need from this list of input options. We'll start with a short
29:39 answer input for the user to write in their first name. And we'll do the same
29:43 for their last name. Notice how we're adding a label above each input, which
29:46 tells the user what to write in each of these. Then, we'll select the email
29:51 block for the user's email. Add in a phone number block. Then, another short answer input for the
29:57 name of their company. Since their budget is a number, we'll add a number
30:01 block to grab that. And finally, we'll add a long answer input where leads can
30:05 describe their specific needs. Now that our form inputs are ready, we can
30:08 customize the form's appearance to match our branding. We can adjust elements
30:14 like the background color, text color, button color, and the accent color. Feel free to
30:21 choose whichever colors you'd like. Now, we can go ahead and publish the form.
30:24 When it's ready, we'll see a sharable link. Let's copy this and actually visit the
30:30 published form. It's looking great and ready to use. In the real world, you
30:33 could either use this as a landing page or embed it into a website. Whenever a
30:38 new lead fills out this form, we need to be storing that response somewhere. For
30:42 our build, we'll store our form response inside Air Table. You can think of Air
30:45 Table kind of like Google Sheets on steroids. It's your database that not
30:50 only stores information, but can process information and if you push it far
30:53 enough, can even be used to build out more complex apps. So sign up for free
30:57 or sign in if you already have an account. On the home screen, you can see
31:02 that we could use AI to build things or start with a template like one for a
31:06 marketing campaign or a project tracker. But for our build, we'll start by
31:10 creating a base from scratch. Base is just Air Table's name for a database.
31:13 Again, it's really similar to a Google sheet or Excel sheet, but can be layered
31:18 with complexity. We'll name it lead base. We'll reference this name later
31:21 when setting up the connections between our tally form and air table and later
31:25 between make and air table. For every field in our form, we need a matching
31:29 field in air table. So, we'll add fields company, budget will reference this to determine
31:56 a lead's qualification. notes, details they submit about their needs created on
32:01 when the lead was added to this base. Importantly, we need to know if our lead
32:05 is qualified or not. Ultimately, lead qualification can be more nuanced, but
32:09 for our needs, we'll qualify leads whose budget is 10,000 plus. By using Air
32:13 Table's AI assistant, we can prompt it to create a custom field for us by
32:17 asking it to create an AI field called qualification that sets the lead as
32:23 qualified or not qualified based on whether the budget field is greater than
32:27 or equal to $10,000. Now, Air Table will use the power of an LLM open AI at the
32:31 time of this recording to help with creating this custom field. After
32:35 processing, the AI fields modal will appear. Since we want leads to be
32:38 autoqualified, we'll enable the automatic generation option.
32:44 If you encounter errors about missing or invalid fields while saving your AI
32:47 prompt, it's typically because the system can't clearly identify which
32:50 fields you're referencing. Double check that all fields are valid and properly
32:56 linked to the correct data in your base. We want our database not just to
33:00 autoqualify leads intelligently, but also to generate a descriptive message
33:04 about each lead that we can share with our sales team. So, we'll use the AI
33:08 assistant again to create a field that generates this message based on the
33:12 lead's information. After double-checking that this new prompt looks good, we can save it and
33:18 move on to the final field. We'll add a date field called contacted on that will
33:23 store the date when we contacted the lead through our automated email. In the
33:26 make workflow, we'll add the date to the field from make at the very end of our
33:30 workflow. Finally, we'll clean up this base by renaming the sheet to lead
33:35 contacts and set up each record or row to be called a lead. Now that we have
33:39 all the necessary Air Table fields that match our form fields, we can connect
33:43 Tally to Air Table via Tally's Air Table Just name the connection, select the
33:52 database and table you're syncing to, and map each of the tally fields to
33:55 their respective data fields in Air After saving this connection, we'll see
34:04 a confirmation that the integration was successful. Now, let's test the
34:07 connection by filling out the form with some dummy data, making sure to set the budget to $10,000
34:19 When we submit, we should see the new lead record populate in our Air Table
34:22 database. And voila, it's showing up perfectly. Since the budget is over
34:26 $10,000, it's automatically marked as qualified. And we've got an AI generated
34:30 message ready to send to our sales team via make. With our lead form synced up
34:33 to our database, we're ready to start building out the automation workflow.
34:37 This is what we're about to build out. We'll have a module that watches our Air
34:41 Table database for new leads. And when it finds one, it sends them an email via
34:45 the Gmail module, then updates that same leads air table row with the time we
34:49 contacted them. It also formats and sends a message into Slack channel. If
34:53 you're not familiar with Slack, it's a professional group chat with channels
34:56 dedicated to specific topics like marketing. With this plan in mind, let's
35:00 head over to make.com. Remember, this is our workflow builder,
35:03 the assembly line where we'll be constructing our automation. You can get
35:07 started for free or login if you already have an account. Once inside, you'll see
35:11 your dashboard, which displays info about your workflows, which are called
35:14 scenarios in Make. You'll see how busy they've been and how much data they've
35:18 used. This helps you monitor your make plan usage, which is particularly
35:21 important if you're on a free plan with usage limits. You can also see if any of
35:25 your scenarios require attention due to things like bugs and errors. The
35:29 scenario tab is where you'll spend most of your time, and you can organize your
35:32 scenarios into folders like you see here. Before we start creating our own,
35:35 I want to bring your attention to the fact that you have a bunch of templates
35:39 to get started from and adapt to your own use cases. And over time, you'll be
35:43 managing the connections here in this tab, which shows you which external apps
35:48 like OpenAI and Google you have connected to from within make. With that
35:52 brief tour out of the way, let's get into it and create a new scenario. We'll
35:56 start by setting up the trigger. After selecting the Air Table module, we'll
36:02 configure it to watch records. Since make doesn't yet know which
36:05 records to watch, we'll need to establish a connection between Make and
36:09 our Air Table account. We'll do so with a token because as you see in this
36:14 warning keys are deprecated. So we just need to create a token and paste it
36:18 here. To get that token go to your Air Table account settings,
36:28 and create a new token named something descriptive like make token.
36:33 When configuring the scope, set it to allow both reading and writing records
36:37 with full database access for read and I'm granting access to all current and
36:45 future bases, though you may want to restrict this depending on your specific
36:49 needs. Now, click create token and immediately copy and store it somewhere
36:52 secure. You'll never be able to see the complete token again. Once we paste the
36:56 token back into make, our connection to air table should be successfully synced
37:01 up. Now we just select the bas and table and set the trigger field to created on.
37:10 The trigger field tells make which field to monitor for changes in order to
37:13 determine when a record is new. By setting it to created on make will only
37:18 pull records whose timestamp is later than the last time it checked for new
37:22 records. That way it only acts on newly added leads and doesn't rerun on old
37:27 ones. As for the label field, we'll set it to the company. This field is just
37:30 for display purposes inside makes scenario logs. It helps you identify the
37:35 records that are being pulled in, but it doesn't change how the automation works.
37:39 The formula field lets you write an air table style formula that filters records
37:43 before they're passed into your automation. Make only processes records
37:47 when this formula evaluates to true. So, if a lead's qualification field is equal
37:52 to qualified, they'll be pushed through to the next step of the automation. If
37:55 they're not qualified, they don't pass this step. If you ever need help writing
37:59 these filters, you can ask an LLM like chat GPT or even have a chat with Make's
38:04 built-in AI. Once we hit save, we'll choose to start from now on. That way,
38:08 make only looks at new leads created after this point and skips anything
38:11 already sitting in Air Table. If we hit the run button, it should successfully
38:15 find the newly created qualified lead and display that records details. Once
38:19 you run this scenario, any qualified leads that were found will not be found
38:23 again on rerun due to this created onfield. You'll have to add a new lead
38:26 each time you want to test this scenario. Again, the automation is
38:31 filtering for only new leads. Since we want to send qualified leads an email to
38:35 schedule a call and notify our sales team about this lead via Slack, we'll
38:39 need to set up two actions. This is where a router comes in. It allows our
38:42 scenario to branch into these two separate tasks simultaneously. Without a
38:47 router, setting up tasks sequentially means if the email step fails, the Slack
38:51 step won't run at all. We want to avoid these waterfall effects where one
38:55 failure blocks other actions. Using a router prevents this issue while making
38:59 our scenario easier to troubleshoot and expand over time. On the first route,
39:03 we'll add the Gmail module and select the send an email action, which requires
39:08 us to set up a connection to our Gmail. If you're using a non-personal email
39:11 like mike@edge.ai, this setup is very simple. You just log into your Gmail account here. But if you
39:18 need to use a personal Gmail, one that actually ends in gmail.com, the process
39:22 is more involved. So, I'm going to ask you to be patient because there are
39:26 quite a few steps involved to connect a personal Gmail account here. If you're
39:29 just watching and not building along yet, feel free to jump past this
39:33 section. Same goes for those of you who are using a non-aggmail.com email. But
39:37 if you do need to connect your personal Gmail account, here's what that looks
39:40 like. As you can see, if we pop open these advanced settings, this requires
39:45 two key pieces of information. a client ID and a client secret. These act as the
39:49 key that allows make to unlock access to your Gmail account. Make provides
39:53 instructions for how to generate these, which you can find by clicking on this
39:56 guide link. At the time of this recording, it links to the help center.
39:59 From here, click into apps documentation, then click on communication and scroll down to Gmail.
40:05 If you're wondering why make didn't just link us to this page, so am I. From
40:09 here, we'll click on create a custom oorthth client. And now we need to log
40:14 into the Google Cloud Platform. Once you log into Google Cloud with the Google
40:17 account you want to send emails from, you'll click select project at the top
40:21 left. From there, click new project, then give the project a name like make and hit
40:29 create. Once it's created, a notification will appear in the top right corner. Click
40:35 select project to open your new project. Now that we're inside, let's enable the
40:41 Gmail API. Go to APIs and services and click enable APIs and services. In
40:45 the search bar, type Gmail and select Then click enable. Next, we need to tell Google who is
40:57 requesting access and what kind of data is being requested. So, we'll click on
41:01 Oorth consent screen and hit get started. To fill out the app
41:05 information, we need to give the app its name. So to keep things consistent,
41:09 let's call it make like we did for the project name earlier. We also need to
41:12 tell it which email we want to give these new abilities to. Now we select
41:17 external for the audience type and next. Under contact information,
41:22 enter an email for Google to notify you about this project. Hit next. Then agree
41:27 to the terms, continue, and hit create. We now need to give make permission to
41:31 interact with Gmail on your behalf. So head into the data access tab and click
41:36 add or remove scopes. We're adding Gmail API scopes. So, search for Gmail and
41:41 then select the following scopes. Both scopes to read, compose, and send
41:46 emails. Manage drafts and send view your emails and settings. View your metadata.
41:51 Add emails to your mailbox. Send email on your behalf. See, and edit labels.
41:55 With all those labels selected, hit update. Then, scroll down and make sure
41:59 to save these scopes. Next, let's configure the branding settings. Scroll
42:03 down to authorize domains and add both make.com and integromat.com which is the
42:08 old name make used to be called. Then save. After that head to the audience
42:13 section and click add users where you'll add your Gmail address and
42:18 hit save. Now it's finally time to create the actual credentials that make will use.
42:23 Go to the client section and click create client. For the application type, choose web
42:30 application and name it make. Scroll down to authorize redirect URIs.
42:35 Click add URIs and paste the exact redirect URI provided by Make. You can
42:39 find this in their Gmail integration Click create and you'll be given your
42:53 You will copy both of these and paste them into the field in makes Gmail
43:01 Now we can click sign in with Google. This sign-in window will appear. Select
43:05 your Gmail account. Grant the requested permissions and hit continue. Once
43:09 that's done, your Gmail account should be successfully connected to make. I
43:13 know that was a lot and I wish it was simpler too. But the good news is we can
43:17 now start sending emails from that Gmail account. So, let's now create the email
43:21 template. We'll set up a subject line using the lead's name from their Air
43:25 Table record and whatever engaging text you want to add in here. For the content
43:29 of the email itself, we can address them by name and say, "We received the
43:32 request about their company and would love to discuss their goals on a call
43:38 they can book via our scheduling link. For this, we will be using a handy tool
43:42 called Kalanley. If you're not familiar, Calendarly is a scheduling tool that
43:45 lets people book meetings with you. You connect it to your calendar, set your
43:49 availability, add conferencing tools like Zoom or Google Meet, then create
43:53 event types like an introductory call and share a link for people to book
43:56 these meetings with you. So just set up a new account if you don't have one or
44:02 log in and we'll create a new event type and call it intro call. This event will
44:06 have a 30-inut duration and the location will be Google Meet. Keep in mind you
44:09 may have to connect a conferencing tool if you're using something like Zoom. Now
44:13 we can set our availability for this intro call event. I already have mine
44:17 set up, but to do that, you just head to the availability tab and go to calendar
44:21 settings. Here you can see I already have a couple calendars synced up, but
44:25 to add a new one, click connect to calendar account and select your
44:28 provider, such as Google. Then just choose the account you want to sync and
44:32 log in. Again, I've already set this up, so I'll close out of this and head back
44:37 to the event. For each event, we can set the availability for that event type. I
44:40 may want to change the time I'm free on Friday, either a specific day or every
44:45 Friday. When I apply these changes, that will update the availability for just
44:49 this event type and does not apply to other events I may have set up in
44:52 Calendarly. Down in more options, I can do things like add an event description
44:56 where I can tell the person booking that I'm excited for our chat. Once we hit
45:00 save changes, the event is ready to share. So, let's copy this link and head
45:05 back into make and paste it into the body of the email we're sending to our
45:08 lead. Finally, we of course need to add the recipient of this email. So, we'll
45:11 fill that in with our lead's email address. With that set up, we can move
45:15 on to our other route below. Ultimately, we're going to be posting a message to a
45:19 Slack channel about our new lead. But if you don't use Slack and don't want to go
45:22 through the steps of connecting another external tool, you could sub this out
45:26 for another Gmail module where we send an email to our sales team letting them
45:29 know that we have a new qualified lead and give them the lead's name and other
45:34 relevant info. Then set the recipient email to whichever test email here. But since I
45:40 want to show you how to use a bunch of popular tools, let's look at setting
45:43 this up to post to Slack. We'll choose the Slack app and select the create a
45:48 message action. If you already have a Slack organization, you can simply
45:56 if you don't, let's quickly walk through creating a new one. You'll head to and
46:00 sign in with a new account, then click create a workspace. Give it a name.
46:04 Here, I'm just adding the name of our mock agency that we're qualifying leads
46:07 for. Then just run through these steps to confirm your name. Skip adding new
46:11 members for now and start with the free version. Now we're in our new Slack
46:15 organization. As you can see, there are channels here for different topics and
46:19 direct messages down here. We'll add a new channel called marketing because
46:22 this is where we'll be posting our Slack message into for the entire sales team
46:26 of our imaginary agency to see. With these ready to go, we can head back to
46:30 make and connect the Slack module to our We'll tell the module where to post the
46:38 message by selecting from a list of our public channels, specifying the
46:42 marketing channel we just created. Now, what should it post to this channel? It
46:45 should post the AI message from our Air Table lead. So, we'll add that in here.
46:49 If we run the scenario, it will work, but the message will look messy because
46:53 it's technically a collection, which just means it includes extra stuff that
46:57 we don't want or need to display here. We can solve this by inserting a
47:00 formatter module just before we post to Slack. We'll select a text passer, which
47:05 will do a pattern match using a regular expression to pull just the clean
47:08 message text out of it. While this might look complicated, think of it like a
47:12 smart highlighter that scans the text and grabs only the part we care about,
47:15 the messages value itself and not the messy metadata. If you ever need to use
47:19 a regular expression or rejects in the future, just ask an LLM to draft one for
47:24 your use case. Now, we tell our text passer which text to pass, the AI
47:28 message. With that formatter taken care of, we can head back into the Slack app
47:32 and tell it to create its message based off that freshly passed text instead of
47:38 the original messy version. As a final step, we want to circle back
47:42 to Air Table and update our lead record with information about when we auto
47:46 contacted them. After the Gmail module, we'll add an Air Table module and use
47:50 the update a record action. We'll configure it to return to the same lead
47:54 base and contacts table targeting the ID of the record that just pushed through
47:57 this scenario. We'll set the contacted onfield to the current date and time by using the now
48:03 expression which simply tells make to insert the exact moment the automation
48:07 runs. If we run the scenario and head back into air table, we'll see the date
48:11 was successfully added to the contacted on field giving our sales team context
48:16 about when the lead was auto that email. Our scenario is all set up and working
48:20 great, but up until now, we've only been running it manually by clicking the run
48:24 once button by toggling on the schedule. We can configure our scenario to run on
48:29 autopilot at regular intervals such as every 15 minutes or on a specific day of
48:33 the week or month. You can even set a custom schedule using advanced
48:36 scheduling using time ranges with start and end dates if you only want it to run
48:40 during a certain window. Once you're confident it's working, just flip that
48:44 schedule toggle on and your automation will run in the background while you
48:47 focus on more important things. Of course, you'll want to frequently save
48:51 the scenario. And if you ever need to revert to a previous version, you can
48:55 revert to that version from here. Scenario inputs are useful for more
48:59 advanced use cases, like when one scenario's output becomes another
49:03 scenario's input, but that's beyond the scope of this basic build. You've also
49:06 got your scenario settings, a place for notes, an auto aligner if your workspace
49:11 gets a little messy, and even a little Finally, there's a quick reference to
49:19 every app and module used in your scenario, helpful for getting a bird's
49:23 eye view of your automation. And so, that's build one explained, just a very
49:27 basic AI qualification based out of Air Table. But to give you an idea of where
49:29 we're going with this, in the next build, build two, we're going to be
49:32 adding in a voice agent here. You may have heard about voice agents before,
49:35 but they're are a really a really exciting uh area of the AI space right
49:38 now. You have two main types. You have inbound and outbound. And uh inbound is
49:42 when you can set up a phone number and people can call. When people call that
49:46 number, then they get to talk to the AI directly. Outbound voice agents is what
49:50 we're going to be building here where we can initiate a call using our automation
49:55 here uh to send a call out to someone and say my phone starts ringing and I
49:58 can pick it up and I'm talking to the AI that we've created here. So this is
50:01 really really cool stuff and super powerful. But the reason we're doing
50:04 this is because the as you will have seen on the form that we set up, we have
50:08 quite a limited amount of information that we're collecting from the business
50:10 and this is for good reason. You don't want to put too much information here uh
50:14 or it will decrease the number of people who fill it out. So sort of having a
50:17 lean form and then if we initially qualify them here in the second build
50:20 we're going to expand it so that we can actually once they are qualified as
50:24 we've explained here the AI and air table is going to determine
50:27 preliminarily if they are qualified then we will use our voice agent to actually
50:31 call the person and ask for more information and we'll walk through a
50:34 script basically and ask them hey tell me a bit about your business and your
50:36 needs what are you hoping to get out of us da da da da basically collecting a
50:40 bunch more information that can then be used for even greater and more accurate
50:44 qualification. So here we have our Vappy voice agent which is going to call them
50:47 and then we're going to get the data of that call, the transcription of that
50:50 call and Vap is actually going to be able to analyze it for us to determine
50:54 if it was a successful qualification or not. And then we have to build the
50:56 automation to handle a few different cases because of course not everyone's
50:59 going to pick up the phone. So here we create a route that handles if they
51:02 answered the phone and then based off the information were they interested or
51:05 were they not interested and if they didn't answer we have some other things
51:08 that we can do to handle it here and update the air table and things like
51:11 that. So, it's essentially build two is adding on top of what we've already
51:15 built on build one uh where we have just a basic qualification and sort of send
51:19 them an email to book in a call or to let the sales team know. Here, we're
51:23 trying to do even deeper qualification and making sure that our sales team
51:26 really isn't getting on any calls they shouldn't be by sending out a voice
51:29 agent to collect more information for us to do an AI pre-qualification for this
51:32 business's lead. So, let's take this automation to the next level by adding
9:21 The AI Automation Landscape
9:24 Now, before we dive deeper, it's important to understand that AI
9:27 automation is a super broad term these days, covering wide ranges of different
9:31 systems and applications that can be built with AI. This is largely due to
9:34 the rapid advances in areas like AI agents and AI tools. So, over the past 2
9:39 years, as I built my own AI agency, Morningside AI, and helped thousands
9:42 through my communities to do the same, I've had to create a clear system for
9:46 making sense of the chaos that is the AI automation landscape. And I really want
9:49 to share that with you today because it's been super helpful for me and many
9:52 of my students to be able to figure this space out and at least have some mental
9:55 buckets that you can put things in. Here's the three different categories
9:57 that you need to keep in mind. And please stick with me. This will all make
10:00 sense in a second. So firstly, we have conversational AI. These are systems
10:03 that chat with people handing back and forth conversations. It's kind of like
10:07 having a friendly robot that talks to customers for you. These kinds of chat
10:09 bots can be found on things like websites and answer questions or they
10:13 can be voice agents that pick up phone calls. These used to need real people to
10:16 talk, but now AI can automate these kinds of conversations. The second
10:19 category is AI tools, and these are systems that use AI to do a specific job
10:23 when a person asks them to, and it's mostly to help workers get more done.
10:27 For example, I can make a custom AI tool that takes a link to a cool blog post
10:31 that I found, grabs the info from the web page, does extra searches on the
10:34 topic, and then uses something like chatbt to write a new beta version for
10:39 my own blog. And third and final is AI workflow automations. These are systems
10:43 that do a whole series of tasks by themselves, starting when something
10:46 happens, like a trigger or on a set schedule like once a day. They use AI to
10:50 make decisions that used to need a human brain. It's kind of like having a smart
10:53 robot manager that runs the whole process for you. For example, an
10:56 automation can call customers of an online store 14 days after they buy
11:01 something using an AI voice agent to ask for feedback and review all without you
11:04 having to do a thing. So, you set it up on that trigger of 14 days after
11:08 purchase, then execute this workflow. Now, those three categories might be a
11:10 little bit confusing if you are completely new to the space, but don't
11:13 worry. In the building section of this video, we're going to be creating three
11:16 automations which integrate each of these different types, and you get to
11:19 see them in action, which will make it super clear. So, AI automation is
11:22 essentially an umbrella term under which all of the exciting stuff in the AI
11:25 space is happening right now. However, when people talk about AI automations,
11:28 they are typically referring to the last type that I mentioned. So, AI workflow
11:31 automation, and these are what we're going to be building later in the video.
11:35 And an automation refers to one chain of steps that uses AI in various ways to do
11:39 certain tasks. And if I'm honest, this last category of AI workflow automation
11:42 is really the most powerful because it can incorporate all elements of AI
11:46 automation like agents, conversational AI, and tools in order to build
11:50 end-to-end processes that are much more valuable than if they were just alone.
11:52 So, what I'm saying is that what you're about to learn is the key skill that
11:55 lets you do pretty much anything in the AI space these days. It is the
11:59 foundation for building systems that save time, make money, or just make life
12:03 easier, whether you're at school or work or running your own show. And after
12:05 years in the game, I can tell you that it's been one of the most valuable
12:10 skills that I have ever picked up. So, now that you understand what AI
12:13 automations are, let's take a little peek under the hood and see how they
12:16 actually work. So, don't worry if this sounds tricky. I've been breaking down
12:19 this kind of complex AI stuff for years now. So, I'm going to make this super
12:21 easy to understand for you. So, you can think of an AI automation like a facto's
12:25 assembly line, right? There's different stations and they're all working
12:28 together to build something awesome from start to finish. It's kind of like
12:31 having a a team of little robots, each with a special job passing the project
12:35 along until it's done. So, let's go through the five key parts to make this
12:38 magic happen. Firstly, we have the trigger. This is the very first step of
12:41 an automation. You can think of it as the facto's start button or the whistle
12:45 that says, "Let's go." It's what kicks everything into gear. It could be
12:48 something like a new email popping into your inbox, a form being filled out on a
12:53 website, or even a specific time of day. This is more of a a schedule. Secondly,
12:57 we have a filter. So, not everything that starts in the automation should
13:00 keep going through it. So, a filter essentially checks if what came in is
13:03 the right stuff to work on. It's like how a factory worker does some kind of
13:06 quality control and checks that if the materials that they've received are good
13:10 enough to use in the final product. If they are not then they get tossed out
13:12 and if they are then they move forward to the next part of the sequence. You
13:15 can think of it kind of like a bounce at a club where only the important stuff
13:18 and the good things that you want inside the club or in your automation are
13:21 allowed through. Thirdly, we have actions. So this is where the real work
13:25 gets done. Actions are the steps your automation takes like the different
13:28 stations in a factory where each one does a specific job. So for example,
13:31 your automation might send an email, update a list or create some kind of
13:34 report. Often there are going to be several actions one after another just
13:37 like a product moving down the line getting built bit by bit in order to
13:41 achieve one of these outcomes. Next we have the intelligence layer. So this is
13:44 where the AI magic shines. This part is like having a super smart robot on the
13:48 assembly line that can think, analyze, and make decisions on the spot. And you
13:51 can tell it how to think using prompting. The AI inside your automation
13:55 can look at each task, figure out what's needed and adapt based on the context
13:58 you provided, like deciding how urgent something is, writing a custom message,
14:02 or pulling out key info from a big mess of data. These intelligent steps can go
14:06 way beyond just following some kind of preset rules that we saw with old school
14:09 automations. Next, we have the format. So, just like items in production line
14:13 may need some sanding or trimming before another piece can be added on top, the
14:16 data in our automations often need some kind of adjustments along the way. And
14:19 this is where we use a formatter to prepare things for the next step. And
14:22 finally, we have the output. This is your finished product, just how Carac
14:25 Factory packages up the final item and ships it out. Your automation delivers
14:28 the completed work at the end of the sequence. This could be a message sent
14:32 to your team, an updated file in your system, or a finished document that's
14:35 ready to go. At the end of the line, it's where everything comes together.
14:38 It's like after making pizza and putting all the toppings on and prepping the
14:40 dough, you're finally getting out of the oven. It's hot. It's ready to eat and
14:43 ready to go. So, let me show you how all of these parts work together with a real
14:46 example that anyone in a job can relate to. Say you're an employee who wants an
14:50 automation to handle incoming customer emails while at work. So the trigger is
14:53 going to be when a new email lands in your company's support inbox. The filter
14:56 is going to check if it's something important like if the email mentions
15:00 urgent or problem then it will pass it along to the next step. If not, it will
15:03 end the sequence there. Thirdly, we have the intelligence layer which is going to
15:06 use AI to read the email, figure out what it's about and then helps to draft
15:10 a helpful response. And in this case, the intelligence layer also acts as a
15:13 formatter which is essentially packaging the response and making sure it's in the
15:16 format we want. Then the actions in the automation can actually send that reply
15:20 and alert the boss on Slack if needed. The output logs everything neatly in
15:24 your system and marks it as handled. So that's just a simple example, but these
15:27 same building blocks are used to automate much bigger things too, which
15:31 is why AI automations are such a game changer these days. They combine these
15:34 parts to create truly smart systems that can save time and boost efficiency,
15:38 whether you are you're juggling tasks at work or crushing it in school or running
15:40 your own business. So, now that you've got a handle on what AI automations are
15:44 and how they're built, let's take a quick look at the tools that make all of
15:48 this possible. Now, don't worry if this sounds a little bit too techy. Soon, these tools are
15:53 going to literally be second nature for you, and it's going to feel so easy to
15:56 do. Creating automation starts with picking a main automation platform.
15:59 These platforms are typically called workflow builders, and they are the
16:03 command center of your automation factory. They give you a blank canvas to
16:06 design your automation on using the building blocks we talked about. Some of
16:09 which are going to be powered by AI like chatb. Popular workflow builders include
16:13 make.com which we're going to be using later in this video in the tutorial
16:16 section. We have Zapia which is great for quick setups and we have NAM which
16:19 is perfect if you want a bit more control. They are the brain of your
16:22 operation and they're really just controlling how everything fits
16:25 together. So workflow builders don't work alone. They are essentially a place
16:28 to hook up all sorts of other tools to get the job done. So these are the
16:30 categories of tools which you can connect to your workflow builders.
16:33 Firstly, we have databases and spreadsheets. So for storing information
16:36 and data, you can use things like Air Table or Google Sheets. You can think of
16:39 them as basically the filing cabinets where you keep all your data neat and
16:42 tidy so that you can save new things to the database or you can pull it into
16:45 your automations as needed. Secondly, we have communication tools for sending
16:49 messages. Things like Slack or Google. They're like kind of walkie-talkies that
16:51 can pass information around automatically for you. Then we have AI
16:55 models. And these can add that smart human level thinking like open AI's chat
16:58 GPT which you can think of kind of like a genius buddy who solves problems for
17:02 you when you give him the instructions that he needs. Then we have scheduling
17:05 tools which can handle time and meetings things like calendarly or Google
17:07 calendar and these are essentially like your personal planner to keep things
17:10 running on schedule. Then we have forms and intake too. So this can collect
17:13 information from people where you have things like type form or tally. These
17:17 are essentially input forms and triggers for your automations when someone fills
17:19 them out. So as your automation goes through it's like calling up all your
17:22 friends on a group project. Each one is bringing their own special knowledge or
17:26 capabilities to the table in order to help you to get to the end of that
17:28 sequence and finish off the automation to create the final product. By building
17:31 out a workflow, you are basically the boss at the factory. Once you know what
17:35 each tool can do, you can mix and match them to work together smoothly along
17:38 this assembly line. It's kind of like building with Lego blocks. You can just
17:42 like snap the right pieces together to make something awesome. You get to
17:45 decide which tools you connect in what order and how AI can make the whole
17:49 system smarter as well. Which AI models do you use? Gemini, do you use OpenAI?
17:52 to use complexity for searching. There's so many different decisions you have to
17:55 make. Uh but it's really cool to be able to flex your kind of creativity in order
17:58 to solve these problems. It's really like a a new age form of problem
18:01 solving, which is why I love it so much. I say, I know I need to take this and
18:05 get to this. How can I use AI to do that? And really forces you to explore
18:08 what's out there in terms of AI tools these days. So, what kinds of workflows
18:11 can we build? So for students, for example, you could build a study
18:14 material organizer that automatically summarizes lecture recordings for you,
18:18 creates flashcards from notes, and schedules review sessions based on your
18:21 exam dates. An employee could improve their workflow with a meeting assistant
18:24 that records and transcribes their meetings, and then generates action item
18:28 summaries off that and updates project management tools. Entrepreneurs could
18:31 design a lead qualification system, which automatically qualifies leads,
18:35 calls them with an AI voice agent, and then sends them a custom proposal to
18:38 automate that whole process. And that is exactly what you'll learn how to build
18:41 by following along with me in this next section. So, I'm going to show you how
18:43 to build out a lead qualification automation step by step, starting simple
18:48 and getting more complex as we go. We're going to be focusing on this
18:50 business-based workflow because these kinds of business systems offer the best
18:54 opportunity for monetizing your new skills, which I'll go into depth on at
18:57 the very end. So, stick with me and you're going to be learning how you can
18:59 start to make money immediately with these kinds of skills. So, before we get into that, let's do a
19:05 quick summary of this section. So firstly, an AI automation is a system
19:09 that uses AI to automatically perform complex tasks that used to require
19:14 humans. Secondly, we build them inside of workflow builders that incorporate
19:17 integrations with tons and tons of other tools. And every automation has six key
19:21 components. Firstly, a trigger, what starts the workflow, a filter,
19:25 conditions that need to be met, the intelligence layer, which processes info
19:29 and makes decisions with AI, the actions, which are the tasks that
19:32 actually get performed, formatterers, which clean things up, and the output.
19:36 the final result or deliverable. So, if you're feeling unclear about anything
19:39 we've covered so far, feel free to go back and rewatch some of those sections
19:42 and be ready to join us in the next step when we begin building. So, it's really,
19:45 really important that you do understand everything that we've gone over there
19:48 because you're not going to have a solid foundation to build your technical
19:51 skills on top of, which we're doing in the next section. So, please, please,
19:54 please, I've taken a lot of time to put all of this information very, very
19:57 gradually together. So, you must understand everything that I've gone
20:00 over just here before we go into the next phase. If that's all good, then
20:02 let's take a look at what we're going to Okay, so now that we have the
20:08 foundational knowledge built that you need, we can now get into the second
20:11 chapter of this video where we're going to be building three AI automations from
20:14 scratch. We're going to be starting off with something super beginner friendly
20:17 to get you started and then working our way up to a much more advanced and
20:20 valuable one by the end. Now, very important note is that each of these
20:23 automations build on each other. So, you have to be able to do the first one in
20:26 order to be able to make the second and so on. So, you you cannot skip ahead.
20:30 I've planned this out very carefully to gradually layer on your skills. So, it's
20:33 all very intentional. So, please just trust the process. So, over the next
20:36 chapter, you'll learn almost all of the key skills you need to start building
20:39 your own AI automations from scratch and be able to tap into this enormous
20:42 opportunity that is AI automation. The system that we're going to be building
20:45 gradually over the next three sections is an AI lead qualification and proposal
20:49 generation system for a business. In the first tutorial, we'll be setting up the
20:53 base using AI to automatically qualify leads after they submit a form on their
20:56 website. In the second tutorial, we'll improve that qualification ability by
21:00 implementing an AI voice agent that can call the lead for more information. And
21:03 finally, in the third tutorial, we'll implement an automated proposal
21:06 generator that can instantly create proposals for qualified leads using the
21:09 information collected on the phone call and in the lead form. The whole point of
21:12 this is that as soon as someone's interested as contacting the business,
21:15 they can kick things off with them immediately by getting a proposal in
21:17 their hand. So this process of qualification is a crucial part of
21:21 running any business at scale because at the end of the day, not all people who
21:24 come to a business are going to be a good fit for their services. So for an
21:27 example, someone may run an accountancy business, but they only choose to work
21:30 with doctors. They are specialized in helping doctors with their finances. But
21:33 if a builder fills out their website form, then that lead would not be
21:37 qualified, right? Taking a call with them would be a waste of time because
21:40 they are not a doctor. Therefore, to stop wasting time, almost all businesses
21:43 need some kind of qualification system. And for qualified leads, most businesses
21:47 need to make some kind of proposal for them in order to kind of look at it and
21:50 and see what they're proposing and then agree to that proposal. Since most
21:54 proposals don't lead to a deal, maybe 20% if you're lucky, this is a huge
21:57 waste of time and resources for the company. So, long story short, what
22:00 we're about to build will solve a number of key problems for basically any
22:03 business, making it extremely valuable for you to be learning how to do it. And
22:05 in fact, this is something you'll be able to go and sell directly to
22:08 businesses when you're done. Here's what this process looks like without AI
22:11 automation first. So, a human sales rep must constantly check for new form
22:14 submissions from the website, review each lead's details, and evaluate if
22:18 they're worth pursuing, research the company to understand how to help them,
22:21 make phone calls, and deliver competent pitches, and then manually create custom
22:25 proposals. That is hours of repetitive work per week. And it could lead to
22:27 potential leads slipping through the cracks. If maybe they get the
22:30 qualification wrong, they don't understand what the business is doing,
22:32 they don't properly research it, or they're just too slow at getting back.
22:35 They may have to wait a whole day for the lead to hear anything back. So
22:37 decreasing the time it's going to take for a qualified lead to hear back from
22:41 them. It's going to drastically increase their conversions. With AI automation,
22:44 we can transform this tedious process into something much more efficient,
22:46 which is exactly what we're going to be building together. So here's how it
22:49 works. It's going to start as usual with a lead filling out the form on the
22:52 website. Then the system is going to automatically qualify them using AI.
22:56 Then it's going to further research the company using AI also. Then it's going
23:00 to send an automated phone call to pitch your offer using an outbound AI voice
23:04 agent. The system then saves the call outcome, summarizes the conversation,
23:07 and generates a personalized proposal for them, all without a human having to
23:11 lift a single finger and in a super scalable way. This is the power of AI
23:15 automation. We're taking tasks that used to demand hours of manual work and
23:18 turning them into workflows that run all by themselves. and we're going to build
23:21 it all using several of the most popular tools from the automation ecosystem. So,
23:24 if you're as excited to learn this as I'm to teach you, then let's get
23:27 started. And everything that you need to follow along with it, including all
23:30 resources, templates, prompts, etc., is going to be available for free on my
23:33 school community. You can find it in the first link in the description. You'll
23:36 need to request to join. It will take 1 to 2 minutes to be accepted. Once you're
23:39 in, you can just search for the title of this video, and then you'll be able to
23:42 find all the resources attached to it. So, that is how you can get all the
23:44 resources to follow along. Let's get stuck into it. All righty, guys. Just to clarify things
23:49 before we jump in, it's very important that you understand what we're doing in
23:52 this first build, but also in the second and the third and how it all fits
23:55 together and you get kind of an idea of where we're going. Understanding this
23:57 technology and then being thrown in a whole bunch of random things uh like Air
24:01 Table and and Slack and make and all of these different terms can be a bit
24:04 confusing. So, I just want very quickly to give you a bit of an orientation
24:07 before we jump into things of what we're building and why and how it works
24:09 because I've touched on it before, but I don't think I did a good enough job. So,
24:12 I'm filming this after just to really make sure you guys are fully clear on
24:15 what we're building here. So, and why it's valuable really as well. So um as
24:20 we touched on this is going to be an AI qualification system um for inbound
24:24 leads to a business. So easiest way to explain this is if we had a a website
24:28 and say we had this book a free consultation form and we might be
24:31 running ads to it. We might just have that on a website and people just
24:34 discover it by searching the web or or we're showing up in Google results. But
24:37 at some point, a business like the example I gave before, if you're an
24:41 accountant and you only work with uh builders or plumbers or whatever the the
24:44 example I gave, some people are going to fill out this form who are not
24:47 necessarily the right person for you. Like the my accountancy firm that only
24:51 works with plumbers, if a surgeon comes in or if a a pilot comes in and fills
24:54 this thing out and wants to get accounty services, we would ask them questions
24:57 around, okay, well, what's your name? What what do you do? Uh what sort of
25:00 industry do you work in? And we'll collect this information here. And as I
25:04 mentioned before, not all of the people who fill out your forms, particular if
25:07 you're running ads or or doing content and it's driving traffic to this landing
25:10 page as a business, not all of them are going to be people that you actually
25:13 want to get on calls with because if you get on calls with them and they're not
25:16 your ICP or the person that you could potentially work with, then it's going
25:19 to be a complete waste of time. So, a qualification is a very essential part
25:22 in in pretty much any business. And what you're going to build is a very very
25:24 powerful system starting off with something basic like this. And then
25:27 we're going to progress it to something a bit more advanced like this. And then
25:30 finally, something extremely valuable and powerful that you guys are all going
25:33 to be able to build along and ultimately sell this to someone if you wanted to
25:35 afterwards. So that's going to be covered in the monetization section. But
25:39 just to clarify sort of what we're going to be doing here in this first build, uh
25:43 you'll start off with building a form on a telly. So we're essentially
25:45 replicating this kind of website form, but we're just going to build it in
25:48 tally in this case. You can embed these into different websites, etc. But just
25:51 think of this as a form that people will get sent to. So maybe they clicked on an
25:54 ad and then they get sent to this. How can we help? We're going to build out
25:57 this form and then when people click submit on this form, it's going to send
26:00 them into this basically a fancy spreadsheet. And in this case, we're
26:03 using Air Table. Air Table is like a really awesome platform which you're
26:05 going to get to to use a lot within this build. Uh but Air Table, you can think
26:09 of it like a fancy spreadsheet and we're going to be able to map uh each of these
26:13 answers to a row in this database here. So every time someone fills this form
26:16 out that we create, it's going to load it and add it as a row into this
26:20 database. And the cool thing there is that air tableable allows us to within
26:24 make.com our workflow builder it allows us to set up events that when things
26:27 happen in air table like okay let's watch for new records that arrive and
26:31 that's what this here is doing. So this trigger of our automation is going to be
26:35 watching for new records that arrive in this and then it's going to take
26:39 anything that arrives in here say this first row and then it's going to put it
26:42 through this automation here. And so this is going to be a very basic one
26:45 where we've taken actually the AI component out of this. Before Air Table
26:48 really got advanced with the AI features. What you're doing here is take
26:51 the information from the Air Table. You'd use some chat GPT step here and
26:54 then you'd be able to do things afterwards. But in this case, I'm making
26:57 it extra simple for you guys and that we're actually just using Air Table's
27:00 built-in AI features. So down here, you can see in the tutorial that you're
27:03 about to go into, we will use Air Table to create an AI column here that's going
27:07 to take in information like uh like here where they're talking about their their
27:10 company, their budget, and about their needs. We can write a prompt in here and
27:13 it's going to automatically fill out this row. So as soon as a new person
27:17 arrives in this spreadsheet, this field that we're creating here is going to
27:20 automatically qualify them and use AI and this prompt that we write here to
27:23 analyze the information that they gave us and then output if they are qualified
27:27 or if they're not. So we're doing the qualification with AI built into Air
27:30 Table which is going to save us a lot of time and messing around with jumping
27:33 back and forth. So So that is basically the start of build one. We will do
27:37 immediate qualification with AI within Air Table and then we're going to if
27:40 they are qualified we're going to send an email to them and because we're
27:43 already doing the qualification here in Air Table it makes it a lot easier for
27:46 us over at make.com because for this particular node as you'll see in a
27:49 second we're going to set it up so that it's only actually going to look for uh
27:54 the rows or records in this table that have been AI qualified as qualified. So
27:58 if this prompt here that we've written analyzes all the information in the row
28:02 and then outputs qualified then and only then will it pass it into here. And this
28:05 is when we're going to be sending an email to the prospect and saying, "Hey,
28:09 we're interested in in hopping on a call. Here's a link to book in a call
28:12 with us." But at the same time, it'll also send a message to our sales team in
28:15 Slack and say, "Hey, look, we've just had a new qualified lead." So, this is
28:19 very much a a minimized version of what you could do uh maybe a year ago. But
28:24 because of the Air Tables AI features, uh we can do a lot of the heavy lifting
28:27 here and writing a prompt that's going to analyze what the person has filled
28:30 out here, aka are you a are you a builder or are you a pilot? And if they
28:34 were a builder, in the case of the example I just gave, it's we're going to
28:36 be using a different example in the build. Um, but it would come out as
28:39 saying qualified here. And because it says qualified, then it would
28:43 automatically trigger uh the rest of this automation, which is sending an
28:45 email to them and taking the next step and saying, "Yes, we are interested in
28:48 talking to you further, making sure that we're not talking to anyone that we
28:51 shouldn't be talking to and at the same time letting the sales team know." So,
28:54 that is build one. Um, that is how it all fits together. And I'm going to be
28:57 doing these little updates just before each build so that you guys are 100%
29:05 To get started, we need a way to gather our leads information. While there are
29:09 several great options for this build, we'll be using Tally, an easy to use
29:13 form builder. Just sign up for free or Then we'll click on new form and give
29:20 our form a title. Anyone using the form will see this title. So, we want it to
29:24 make sense for the user. So, in our case, we can name it, how can we help? I
29:27 want to build it from scratch. So, I'll hit enter. The way we create forms in
29:31 tally is with building blocks. So, if we click the plus button, we can select the
29:35 type of block we need from this list of input options. We'll start with a short
29:39 answer input for the user to write in their first name. And we'll do the same
29:43 for their last name. Notice how we're adding a label above each input, which
29:46 tells the user what to write in each of these. Then, we'll select the email
29:51 block for the user's email. Add in a phone number block. Then, another short answer input for the
29:57 name of their company. Since their budget is a number, we'll add a number
30:01 block to grab that. And finally, we'll add a long answer input where leads can
30:05 describe their specific needs. Now that our form inputs are ready, we can
30:08 customize the form's appearance to match our branding. We can adjust elements
30:14 like the background color, text color, button color, and the accent color. Feel free to
30:21 choose whichever colors you'd like. Now, we can go ahead and publish the form.
30:24 When it's ready, we'll see a sharable link. Let's copy this and actually visit the
30:30 published form. It's looking great and ready to use. In the real world, you
30:33 could either use this as a landing page or embed it into a website. Whenever a
30:38 new lead fills out this form, we need to be storing that response somewhere. For
30:42 our build, we'll store our form response inside Air Table. You can think of Air
30:45 Table kind of like Google Sheets on steroids. It's your database that not
30:50 only stores information, but can process information and if you push it far
30:53 enough, can even be used to build out more complex apps. So sign up for free
30:57 or sign in if you already have an account. On the home screen, you can see
31:02 that we could use AI to build things or start with a template like one for a
31:06 marketing campaign or a project tracker. But for our build, we'll start by
31:10 creating a base from scratch. Base is just Air Table's name for a database.
31:13 Again, it's really similar to a Google sheet or Excel sheet, but can be layered
31:18 with complexity. We'll name it lead base. We'll reference this name later
31:21 when setting up the connections between our tally form and air table and later
31:25 between make and air table. For every field in our form, we need a matching
31:29 field in air table. So, we'll add fields company, budget will reference this to determine
31:56 a lead's qualification. notes, details they submit about their needs created on
32:01 when the lead was added to this base. Importantly, we need to know if our lead
32:05 is qualified or not. Ultimately, lead qualification can be more nuanced, but
32:09 for our needs, we'll qualify leads whose budget is 10,000 plus. By using Air
32:13 Table's AI assistant, we can prompt it to create a custom field for us by
32:17 asking it to create an AI field called qualification that sets the lead as
32:23 qualified or not qualified based on whether the budget field is greater than
32:27 or equal to $10,000. Now, Air Table will use the power of an LLM open AI at the
32:31 time of this recording to help with creating this custom field. After
32:35 processing, the AI fields modal will appear. Since we want leads to be
32:38 autoqualified, we'll enable the automatic generation option.
32:44 If you encounter errors about missing or invalid fields while saving your AI
32:47 prompt, it's typically because the system can't clearly identify which
32:50 fields you're referencing. Double check that all fields are valid and properly
32:56 linked to the correct data in your base. We want our database not just to
33:00 autoqualify leads intelligently, but also to generate a descriptive message
33:04 about each lead that we can share with our sales team. So, we'll use the AI
33:08 assistant again to create a field that generates this message based on the
33:12 lead's information. After double-checking that this new prompt looks good, we can save it and
33:18 move on to the final field. We'll add a date field called contacted on that will
33:23 store the date when we contacted the lead through our automated email. In the
33:26 make workflow, we'll add the date to the field from make at the very end of our
33:30 workflow. Finally, we'll clean up this base by renaming the sheet to lead
33:35 contacts and set up each record or row to be called a lead. Now that we have
33:39 all the necessary Air Table fields that match our form fields, we can connect
33:43 Tally to Air Table via Tally's Air Table Just name the connection, select the
33:52 database and table you're syncing to, and map each of the tally fields to
33:55 their respective data fields in Air After saving this connection, we'll see
34:04 a confirmation that the integration was successful. Now, let's test the
34:07 connection by filling out the form with some dummy data, making sure to set the budget to $10,000
34:19 When we submit, we should see the new lead record populate in our Air Table
34:22 database. And voila, it's showing up perfectly. Since the budget is over
34:26 $10,000, it's automatically marked as qualified. And we've got an AI generated
34:30 message ready to send to our sales team via make. With our lead form synced up
34:33 to our database, we're ready to start building out the automation workflow.
34:37 This is what we're about to build out. We'll have a module that watches our Air
34:41 Table database for new leads. And when it finds one, it sends them an email via
34:45 the Gmail module, then updates that same leads air table row with the time we
34:49 contacted them. It also formats and sends a message into Slack channel. If
34:53 you're not familiar with Slack, it's a professional group chat with channels
34:56 dedicated to specific topics like marketing. With this plan in mind, let's
35:00 head over to make.com. Remember, this is our workflow builder,
35:03 the assembly line where we'll be constructing our automation. You can get
35:07 started for free or login if you already have an account. Once inside, you'll see
35:11 your dashboard, which displays info about your workflows, which are called
35:14 scenarios in Make. You'll see how busy they've been and how much data they've
35:18 used. This helps you monitor your make plan usage, which is particularly
35:21 important if you're on a free plan with usage limits. You can also see if any of
35:25 your scenarios require attention due to things like bugs and errors. The
35:29 scenario tab is where you'll spend most of your time, and you can organize your
35:32 scenarios into folders like you see here. Before we start creating our own,
35:35 I want to bring your attention to the fact that you have a bunch of templates
35:39 to get started from and adapt to your own use cases. And over time, you'll be
35:43 managing the connections here in this tab, which shows you which external apps
35:48 like OpenAI and Google you have connected to from within make. With that
35:52 brief tour out of the way, let's get into it and create a new scenario. We'll
35:56 start by setting up the trigger. After selecting the Air Table module, we'll
36:02 configure it to watch records. Since make doesn't yet know which
36:05 records to watch, we'll need to establish a connection between Make and
36:09 our Air Table account. We'll do so with a token because as you see in this
36:14 warning keys are deprecated. So we just need to create a token and paste it
36:18 here. To get that token go to your Air Table account settings,
36:28 and create a new token named something descriptive like make token.
36:33 When configuring the scope, set it to allow both reading and writing records
36:37 with full database access for read and I'm granting access to all current and
36:45 future bases, though you may want to restrict this depending on your specific
36:49 needs. Now, click create token and immediately copy and store it somewhere
36:52 secure. You'll never be able to see the complete token again. Once we paste the
36:56 token back into make, our connection to air table should be successfully synced
37:01 up. Now we just select the bas and table and set the trigger field to created on.
37:10 The trigger field tells make which field to monitor for changes in order to
37:13 determine when a record is new. By setting it to created on make will only
37:18 pull records whose timestamp is later than the last time it checked for new
37:22 records. That way it only acts on newly added leads and doesn't rerun on old
37:27 ones. As for the label field, we'll set it to the company. This field is just
37:30 for display purposes inside makes scenario logs. It helps you identify the
37:35 records that are being pulled in, but it doesn't change how the automation works.
37:39 The formula field lets you write an air table style formula that filters records
37:43 before they're passed into your automation. Make only processes records
37:47 when this formula evaluates to true. So, if a lead's qualification field is equal
37:52 to qualified, they'll be pushed through to the next step of the automation. If
37:55 they're not qualified, they don't pass this step. If you ever need help writing
37:59 these filters, you can ask an LLM like chat GPT or even have a chat with Make's
38:04 built-in AI. Once we hit save, we'll choose to start from now on. That way,
38:08 make only looks at new leads created after this point and skips anything
38:11 already sitting in Air Table. If we hit the run button, it should successfully
38:15 find the newly created qualified lead and display that records details. Once
38:19 you run this scenario, any qualified leads that were found will not be found
38:23 again on rerun due to this created onfield. You'll have to add a new lead
38:26 each time you want to test this scenario. Again, the automation is
38:31 filtering for only new leads. Since we want to send qualified leads an email to
38:35 schedule a call and notify our sales team about this lead via Slack, we'll
38:39 need to set up two actions. This is where a router comes in. It allows our
38:42 scenario to branch into these two separate tasks simultaneously. Without a
38:47 router, setting up tasks sequentially means if the email step fails, the Slack
38:51 step won't run at all. We want to avoid these waterfall effects where one
38:55 failure blocks other actions. Using a router prevents this issue while making
38:59 our scenario easier to troubleshoot and expand over time. On the first route,
39:03 we'll add the Gmail module and select the send an email action, which requires
39:08 us to set up a connection to our Gmail. If you're using a non-personal email
39:11 like mike@edge.ai, this setup is very simple. You just log into your Gmail account here. But if you
39:18 need to use a personal Gmail, one that actually ends in gmail.com, the process
39:22 is more involved. So, I'm going to ask you to be patient because there are
39:26 quite a few steps involved to connect a personal Gmail account here. If you're
39:29 just watching and not building along yet, feel free to jump past this
39:33 section. Same goes for those of you who are using a non-aggmail.com email. But
39:37 if you do need to connect your personal Gmail account, here's what that looks
39:40 like. As you can see, if we pop open these advanced settings, this requires
39:45 two key pieces of information. a client ID and a client secret. These act as the
39:49 key that allows make to unlock access to your Gmail account. Make provides
39:53 instructions for how to generate these, which you can find by clicking on this
39:56 guide link. At the time of this recording, it links to the help center.
39:59 From here, click into apps documentation, then click on communication and scroll down to Gmail.
40:05 If you're wondering why make didn't just link us to this page, so am I. From
40:09 here, we'll click on create a custom oorthth client. And now we need to log
40:14 into the Google Cloud Platform. Once you log into Google Cloud with the Google
40:17 account you want to send emails from, you'll click select project at the top
40:21 left. From there, click new project, then give the project a name like make and hit
40:29 create. Once it's created, a notification will appear in the top right corner. Click
40:35 select project to open your new project. Now that we're inside, let's enable the
40:41 Gmail API. Go to APIs and services and click enable APIs and services. In
40:45 the search bar, type Gmail and select Then click enable. Next, we need to tell Google who is
40:57 requesting access and what kind of data is being requested. So, we'll click on
41:01 Oorth consent screen and hit get started. To fill out the app
41:05 information, we need to give the app its name. So to keep things consistent,
41:09 let's call it make like we did for the project name earlier. We also need to
41:12 tell it which email we want to give these new abilities to. Now we select
41:17 external for the audience type and next. Under contact information,
41:22 enter an email for Google to notify you about this project. Hit next. Then agree
41:27 to the terms, continue, and hit create. We now need to give make permission to
41:31 interact with Gmail on your behalf. So head into the data access tab and click
41:36 add or remove scopes. We're adding Gmail API scopes. So, search for Gmail and
41:41 then select the following scopes. Both scopes to read, compose, and send
41:46 emails. Manage drafts and send view your emails and settings. View your metadata.
41:51 Add emails to your mailbox. Send email on your behalf. See, and edit labels.
41:55 With all those labels selected, hit update. Then, scroll down and make sure
41:59 to save these scopes. Next, let's configure the branding settings. Scroll
42:03 down to authorize domains and add both make.com and integromat.com which is the
42:08 old name make used to be called. Then save. After that head to the audience
42:13 section and click add users where you'll add your Gmail address and
42:18 hit save. Now it's finally time to create the actual credentials that make will use.
42:23 Go to the client section and click create client. For the application type, choose web
42:30 application and name it make. Scroll down to authorize redirect URIs.
42:35 Click add URIs and paste the exact redirect URI provided by Make. You can
42:39 find this in their Gmail integration Click create and you'll be given your
42:53 You will copy both of these and paste them into the field in makes Gmail
43:01 Now we can click sign in with Google. This sign-in window will appear. Select
43:05 your Gmail account. Grant the requested permissions and hit continue. Once
43:09 that's done, your Gmail account should be successfully connected to make. I
43:13 know that was a lot and I wish it was simpler too. But the good news is we can
43:17 now start sending emails from that Gmail account. So, let's now create the email
43:21 template. We'll set up a subject line using the lead's name from their Air
43:25 Table record and whatever engaging text you want to add in here. For the content
43:29 of the email itself, we can address them by name and say, "We received the
43:32 request about their company and would love to discuss their goals on a call
43:38 they can book via our scheduling link. For this, we will be using a handy tool
43:42 called Kalanley. If you're not familiar, Calendarly is a scheduling tool that
43:45 lets people book meetings with you. You connect it to your calendar, set your
43:49 availability, add conferencing tools like Zoom or Google Meet, then create
43:53 event types like an introductory call and share a link for people to book
43:56 these meetings with you. So just set up a new account if you don't have one or
44:02 log in and we'll create a new event type and call it intro call. This event will
44:06 have a 30-inut duration and the location will be Google Meet. Keep in mind you
44:09 may have to connect a conferencing tool if you're using something like Zoom. Now
44:13 we can set our availability for this intro call event. I already have mine
44:17 set up, but to do that, you just head to the availability tab and go to calendar
44:21 settings. Here you can see I already have a couple calendars synced up, but
44:25 to add a new one, click connect to calendar account and select your
44:28 provider, such as Google. Then just choose the account you want to sync and
44:32 log in. Again, I've already set this up, so I'll close out of this and head back
44:37 to the event. For each event, we can set the availability for that event type. I
44:40 may want to change the time I'm free on Friday, either a specific day or every
44:45 Friday. When I apply these changes, that will update the availability for just
44:49 this event type and does not apply to other events I may have set up in
44:52 Calendarly. Down in more options, I can do things like add an event description
44:56 where I can tell the person booking that I'm excited for our chat. Once we hit
45:00 save changes, the event is ready to share. So, let's copy this link and head
45:05 back into make and paste it into the body of the email we're sending to our
45:08 lead. Finally, we of course need to add the recipient of this email. So, we'll
45:11 fill that in with our lead's email address. With that set up, we can move
45:15 on to our other route below. Ultimately, we're going to be posting a message to a
45:19 Slack channel about our new lead. But if you don't use Slack and don't want to go
45:22 through the steps of connecting another external tool, you could sub this out
45:26 for another Gmail module where we send an email to our sales team letting them
45:29 know that we have a new qualified lead and give them the lead's name and other
45:34 relevant info. Then set the recipient email to whichever test email here. But since I
45:40 want to show you how to use a bunch of popular tools, let's look at setting
45:43 this up to post to Slack. We'll choose the Slack app and select the create a
45:48 message action. If you already have a Slack organization, you can simply
45:56 if you don't, let's quickly walk through creating a new one. You'll head to and
46:00 sign in with a new account, then click create a workspace. Give it a name.
46:04 Here, I'm just adding the name of our mock agency that we're qualifying leads
46:07 for. Then just run through these steps to confirm your name. Skip adding new
46:11 members for now and start with the free version. Now we're in our new Slack
46:15 organization. As you can see, there are channels here for different topics and
46:19 direct messages down here. We'll add a new channel called marketing because
46:22 this is where we'll be posting our Slack message into for the entire sales team
46:26 of our imaginary agency to see. With these ready to go, we can head back to
46:30 make and connect the Slack module to our We'll tell the module where to post the
46:38 message by selecting from a list of our public channels, specifying the
46:42 marketing channel we just created. Now, what should it post to this channel? It
46:45 should post the AI message from our Air Table lead. So, we'll add that in here.
46:49 If we run the scenario, it will work, but the message will look messy because
46:53 it's technically a collection, which just means it includes extra stuff that
46:57 we don't want or need to display here. We can solve this by inserting a
47:00 formatter module just before we post to Slack. We'll select a text passer, which
47:05 will do a pattern match using a regular expression to pull just the clean
47:08 message text out of it. While this might look complicated, think of it like a
47:12 smart highlighter that scans the text and grabs only the part we care about,
47:15 the messages value itself and not the messy metadata. If you ever need to use
47:19 a regular expression or rejects in the future, just ask an LLM to draft one for
47:24 your use case. Now, we tell our text passer which text to pass, the AI
47:28 message. With that formatter taken care of, we can head back into the Slack app
47:32 and tell it to create its message based off that freshly passed text instead of
47:38 the original messy version. As a final step, we want to circle back
47:42 to Air Table and update our lead record with information about when we auto
47:46 contacted them. After the Gmail module, we'll add an Air Table module and use
47:50 the update a record action. We'll configure it to return to the same lead
47:54 base and contacts table targeting the ID of the record that just pushed through
47:57 this scenario. We'll set the contacted onfield to the current date and time by using the now
48:03 expression which simply tells make to insert the exact moment the automation
48:07 runs. If we run the scenario and head back into air table, we'll see the date
48:11 was successfully added to the contacted on field giving our sales team context
48:16 about when the lead was auto that email. Our scenario is all set up and working
48:20 great, but up until now, we've only been running it manually by clicking the run
48:24 once button by toggling on the schedule. We can configure our scenario to run on
48:29 autopilot at regular intervals such as every 15 minutes or on a specific day of
48:33 the week or month. You can even set a custom schedule using advanced
48:36 scheduling using time ranges with start and end dates if you only want it to run
48:40 during a certain window. Once you're confident it's working, just flip that
48:44 schedule toggle on and your automation will run in the background while you
48:47 focus on more important things. Of course, you'll want to frequently save
48:51 the scenario. And if you ever need to revert to a previous version, you can
48:55 revert to that version from here. Scenario inputs are useful for more
48:59 advanced use cases, like when one scenario's output becomes another
49:03 scenario's input, but that's beyond the scope of this basic build. You've also
49:06 got your scenario settings, a place for notes, an auto aligner if your workspace
49:11 gets a little messy, and even a little Finally, there's a quick reference to
49:19 every app and module used in your scenario, helpful for getting a bird's
49:23 eye view of your automation. And so, that's build one explained, just a very
49:27 basic AI qualification based out of Air Table. But to give you an idea of where
49:29 we're going with this, in the next build, build two, we're going to be
49:32 adding in a voice agent here. You may have heard about voice agents before,
49:35 but they're are a really a really exciting uh area of the AI space right
49:38 now. You have two main types. You have inbound and outbound. And uh inbound is
49:42 when you can set up a phone number and people can call. When people call that
49:46 number, then they get to talk to the AI directly. Outbound voice agents is what
49:50 we're going to be building here where we can initiate a call using our automation
49:55 here uh to send a call out to someone and say my phone starts ringing and I
49:58 can pick it up and I'm talking to the AI that we've created here. So this is
50:01 really really cool stuff and super powerful. But the reason we're doing
50:04 this is because the as you will have seen on the form that we set up, we have
50:08 quite a limited amount of information that we're collecting from the business
50:10 and this is for good reason. You don't want to put too much information here uh
50:14 or it will decrease the number of people who fill it out. So sort of having a
50:17 lean form and then if we initially qualify them here in the second build
50:20 we're going to expand it so that we can actually once they are qualified as
50:24 we've explained here the AI and air table is going to determine
50:27 preliminarily if they are qualified then we will use our voice agent to actually
50:31 call the person and ask for more information and we'll walk through a
50:34 script basically and ask them hey tell me a bit about your business and your
50:36 needs what are you hoping to get out of us da da da da basically collecting a
50:40 bunch more information that can then be used for even greater and more accurate
50:44 qualification. So here we have our Vappy voice agent which is going to call them
50:47 and then we're going to get the data of that call, the transcription of that
50:50 call and Vap is actually going to be able to analyze it for us to determine
50:54 if it was a successful qualification or not. And then we have to build the
50:56 automation to handle a few different cases because of course not everyone's
50:59 going to pick up the phone. So here we create a route that handles if they
51:02 answered the phone and then based off the information were they interested or
51:05 were they not interested and if they didn't answer we have some other things
51:08 that we can do to handle it here and update the air table and things like
51:11 that. So, it's essentially build two is adding on top of what we've already
51:15 built on build one uh where we have just a basic qualification and sort of send
51:19 them an email to book in a call or to let the sales team know. Here, we're
51:23 trying to do even deeper qualification and making sure that our sales team
51:26 really isn't getting on any calls they shouldn't be by sending out a voice
51:29 agent to collect more information for us to do an AI pre-qualification for this
51:32 business's lead. So, let's take this automation to the next level by adding
51:36 AI voice calls to qualify leads even more effectively. We've already built a
51:39 strong foundation, a system that identifies qualified leads, follows up
51:44 via email, and notifies your team. This alone puts you ahead of the curve. But
51:47 what if we could make the system even more responsive, more conversational,
51:51 more human even? That's where an AI voice agent comes in. By weaving it into
51:55 our workflow, we can automatically call leads, gauge their interest through real
51:58 conversation, and tailor our follow-up based on their responses, all without
52:02 lifting a finger. This is where automation becomes more than a task
52:05 runner. It becomes a teammate. It's the closest thing we have to a human
52:08 conversation at scale. And when it comes to leads, timing and tone matter.
52:12 Research shows that responding to leads immediately increases success rates by 7
52:16 to 9x. In this next section, we'll integrate VP, a popular AI voice agent
52:20 service, into our workflow. First, we'll set up our voice agent. Then, we'll
52:23 configure make to have it call our leads. Finally, we'll enhance its
52:26 effectiveness by feeding it custom research about each lead, enabling more
52:30 targeted pitches. So, what exactly is an AI powered voice agent? You can think of
52:33 them like a version of Siri or Alexa, but one that's specifically designed to
52:37 have natural phone conversations with people. Depending on your needs, it can
52:40 either make calls for you or answer calls. These agents can do amazing
52:43 things like follow up with potential customers, schedule appointments, answer
52:47 common questions, collect feedback, or conduct surveys. There are several
52:50 platforms that offer AI voice agents like 11 Labs, great for highly realistic
52:55 voices, and Vappy, known for being fast and affordable for our automation. We'll
52:58 be using Vapy because it's beginner friendly, cost-ffective, and integrates
53:02 well with other tools. If you haven't already created an account, go ahead and
53:06 do that first. Once inside, we'll go to the assistance tab and click create assistant. This
53:12 starts the process of configuring our assistant. You can start from a template
53:15 like a lead qualification specialist, but we'll create our own from scratch
53:21 and name it Ben. When choosing which provider and model to use, you'll want to balance ability
53:26 with speed and cost. So, keep the price in mind and the time it takes to
53:30 actually respond, which we call latency, which you can see next to the model
53:33 options. For now, we'll go with the cheapest and fastest option, which is
53:37 GPT4.1 Mini at the time of this recording. The most crucial element is
53:41 the prompt, which defines your agents behavior and objectives during the call.
53:45 First, we specify the opening greeting message that the agent will use when
53:49 calling someone. We'll open up the call asking if it's a good time to talk about
53:53 the lead's business needs. The actual prompt is much more detailed
53:58 and I'll paste in one I wrote earlier. You can find this prompt and everything
54:01 else you'll need to follow along in the first link in the video description.
54:06 Let's take a look at what it includes. As you can see, we start off by defining
54:11 its identity and role. A voice assistant representing Edge AI and AI automation
54:16 agency. then specify its goal to pitch our services and determine whether the lead
54:22 would like to receive a proposal. We then set its tone and behavior. A
54:27 knowledgeable rep who is curious but not pushy and who uses casual natural
54:32 language with words like uh and mhm. This section helps humanize your
54:36 assistant. Finally, we outline the structure of the call that it should
54:40 replicate. The conversation flow starts with introducing our AI services, then
54:44 moves to understanding the prospect's needs through targeted questions. After
54:48 acknowledging their responses, we pivot to offering a proposal. If pricing comes
54:53 up, we defer to the proposal for details. The call ends either by
54:56 confirming they'll send the proposal over email or with a plight goodbye if
55:00 they decline. It's best to outline the steps as clear bullet points like this,
55:04 but keep them focused. Too many steps can overwhelm the AI assistant. In our
55:08 prompt, we included context about the company the assistant is representing.
55:11 For our needs, this is sufficient. But if our call was more involved, like a
55:14 customer support agent who is receiving hundreds of nuanced calls, we could add
55:18 files into its knowledge base that it can reference, like all of our policies
55:22 and procedures. But keep in mind, this will increase the response latency since
55:25 the assistant will need to make round trips to these files as it generates its
55:29 responses. So, if you can get away without adding supplemental files, your
55:32 agent will be more performant. Our assistant is almost ready to start
55:35 making calls on our behalf. But before we publish it, let's add some extra
55:39 configuration. As we scroll down, we'll see that we can tweak the transcriber
55:42 settings, which we'll leave alone for now. We can also select different voices
55:46 depending on the personality we want our agent to have. For our agent, we'll go
55:49 with Vap's Elliot voice, but feel free to try out different options to feel out
55:53 which one is best for your use case. You can also add background noise to the
55:57 call. Adding ambient sound makes conversations feel more natural and
56:01 helps mask any brief delays that occur while the agent generates responses
56:04 instead of awkward silence. Callers will hear realistic background sounds. Next,
56:08 you'll see the tools section, which unlocks powerful ways for your agent to
56:12 take action during a call, like sending data to your CRM, triggering external
56:17 workflows, or running custom logic. This is where Vappy really shines for more
56:21 advanced automations. However, since it's beyond the scope of this tutorial,
56:24 we'll leave it untouched for now. What we will be using is the call summary.
56:28 Vari automatically transcribes the call and the prompt inside here tells it how
56:31 to summarize the call based on that transcript. If we open it up, we'll see
56:35 there's already a built-in prompt that tells the assistant how to create its
56:39 call summary. While this default prompt works decently, we can make it even more
56:42 valuable for our make scenario by customizing it to capture additional
56:46 details. Here, I'm pasting in a new prompt telling it to summarize the call
56:50 in a couple sentences with details that will help out with creating a proposal
56:53 later on. Since we'll add this call summary to our lead's air table record,
56:57 capturing detailed insights here is valuable. We also want to tell Vappy how
57:00 to judge if a call was successful or not. There's already a decent default
57:04 prompt here, but we'll customize it for our needs by telling it to judge success
57:08 solely on whether the lead wants to receive a proposal. We'll evaluate
57:11 success with a pass fail, which simply gives us a true or false value that we
57:14 can use in our automation workflow. While there are additional advanced
57:18 options here, the final step we're concerned about is the voicemail
57:22 detection. Let's make sure we toggle that on with VP as the provider. This is
57:26 helpful for our make scenario since we need to accurately track whether the
57:30 call was actually answered or if it just went to voicemail. We're now ready to
57:34 hit publish and test our newly created voice assistant. Hey, it's Ben from
57:39 Edji. Is now a good time to talk about your business needs? Yes, it is. Great.
57:44 I'm following up on your submission because we build custom AI agents and
57:48 automations for companies like yours. anything from customerf facing bots to
57:52 internal workflow optimization. I'd love to understand a bit more about what led
57:57 you uh to reach out. What sort of processes or tasks are you looking to
58:01 automate? I'm looking to implement custom AI agents to facilitate learning
58:05 on an e-learning platform. That sounds really interesting. Um so you're looking
58:10 um to enhance uh the learning experience on your e-learning platform with AI
58:14 agents. Are there any specific challenges or bottlenecks you're facing
58:18 right now that you hope these agents could help with? Uh, yeah, just
58:21 streamlining the learning of the content library on my platform.
58:26 Got it. So, you're looking to streamline how users access and learn from your
58:30 content library. That makes a lot of sense. Thanks for sharing that. Would it
58:34 be helpful if I sent a quick proposal your way showing how we could support
58:40 your goals? Yeah, that'd be great. Perfect. I'll send that to the email we
58:44 have on file. Thanks again for your time and I look forward to helping you with
58:49 your project. Great. Thanks. Bye. You're welcome. Have a great day. Bye. Great.
58:54 Our assistant is working as intended. It's now the voice of our lead outreach
58:58 standing by, ready to make calls on our behalf. Before integrating this
59:01 assistant inside of Make, let's walk through the new path a lead will take
59:04 through our extended scenario. We left off here, but we're going to extend this
59:08 for the trigger. Nothing changes here. A new Air Table record marked qualified
59:12 still kicks things off. Then our voice agent will call the lead using the
59:15 number from their record. We'll add a short pause to wait for the conversation
59:18 to finish before we analyze what happened. We grab the call record from
59:22 VAP. Then we check if the call was answered. Answered. Great. We log the
59:27 call summary and mark the lead as interested or not in Air Table. No
59:32 answer. No problem. We fall back to emailing them a link to schedule a call
59:35 and alert our sales team. Since we're going to be logging call summaries and
59:39 interest level about our leads, we'll just need to head to our lead base in
59:43 Air Table and add those. We'll add a long text field called summary,
59:48 a checkbox for interested, and while we're at it, let's add a new date for
59:52 when a proposal was sent on since in the final phase of this buildout, we'll be
59:55 generating and sending proposals to these interested leads. Back over in
59:59 make, we're ready to extend the lead qualifier scenario that we built
60:03 earlier. To keep things clean, let's just clone what we built and extend from
60:07 there. Name it lead qualifier plus voice agent. And now we're ready to continue
60:11 building. The beginning step remain the same where we watch for new leads.
60:17 Then we'll add the VP module to create an outbound phone call.
60:22 We will need to set up the connection to our VP account, which means we'll head back into that
60:28 dashboard, click on API keys, then copy your private key. If you don't see one
60:31 in there, just go ahead and add a new one. Give it a relevant name. Decide if
60:35 you want to restrict it to only work on certain sites. Or if you only want it to
60:39 work with specific assistant, then create the private token. You'll see it
60:43 pop up here. Then just copy it. But make sure to save this somewhere safe since
60:47 you won't be able to view it again. Pasting the key into the VP module
60:50 should set up our connection. Then we can configure our call. We'll just fill
60:53 in these fields, giving it the assistance ID, which you copy from the
60:58 top of the Vappy Assistant page. We provide the assistant with the lead's
61:01 phone number pulled from their air table record. But what number is our assistant
61:06 calling from? If we head back into Vappy and click on the phone numbers tab, we
61:09 can create a new free phone number with Vappy, specifying the area code to call
61:13 from. Note that at the time of this recording, only US area codes are
61:17 supported by Vappy. Since we're building this assistant for learning purposes, we
61:23 can simply use a US area code like 223. However, if you plan to deploy an
61:27 assistant for production use and you or your client are located outside the US,
61:30 you'll need to import a number from something like Twilio or Vonnage. Okay.
61:34 Once that number is created, we can copy its ID, not the number itself, and paste
61:38 it into the makevarpy module. And with that, our voice agent is set up and
61:42 ready to make calls triggered by our make scenario. Once the call is sent, we
61:45 need to wait a bit before we check what happened. Calls take time to connect,
61:49 ring, and potentially have a conversation. If we try to fetch the
61:52 result too quickly, we could just get an empty response or an error. So, let's
61:56 add a short pause. Insert a tools sleep module. Set the delay to 300 seconds.
62:01 This little breather, 5 minutes exactly, gives Vappy time to do its thing before
62:05 we peek in to see how it went. To ensure things are working so far, let's add a
62:08 new lead to our Air Table base. We could add them via the form, but it's quicker
62:12 just to duplicate an existing lead in order to create a new one. Now, we'll
62:15 head back and make and run the scenario Hey, it's Ben from Edgeai. Is now a good
62:25 time to talk about your business? Yeah, it is. Great. I'm following up on your
62:32 submission about exploring AI solutions. We build custom AI agents and
62:36 automations for companies like yours. Anything from customerf facing bots to
62:40 internal workflow optimization. I'd love to understand a bit more about what led
62:45 you to reach out. What sort of processes or tasks are you looking to automate?
62:49 Yeah. So, I'm just interested in um seeing how we can build out some maybe
62:53 AI powered workflows to help the That sounds interesting. AI powered
63:02 workflows can really streamline operations. Are there any specific
63:06 bottlenecks or manual tasks you're hoping to reduce? Hoping to help my uh
63:12 learning management system run more Got it. Optimizing your learning
63:19 management system can definitely enhance efficiency and user experience. Uh,
63:24 thanks for sharing that. Um, would it be helpful if I sent a quick proposal your
63:27 way showing how we could support your Perfect. I'll send that to the email we
63:36 have on file. Thanks again and I look forward to your feedback. Thank you.
63:41 You're welcome. Have a great day. Now comes the part where the automation
63:45 becomes observant. We're not just automating calls. We're building a
63:48 system that pays attention to what happened. Did someone pick up? Did the
63:52 assistant get the job done? Was the lead interested? To answer those questions,
63:56 we need to ask VBY for the results of a specific call. And to do that, we'll use
64:00 an HTTP module to make a request for that call record. To better understand
64:04 this module's function, let's explore what an HTTP request is and how it fits
64:08 into the broader workings of the internet and how APIs fit into all of
64:13 this. I know it sounds complicated, but let's break it down with a simple
64:16 analogy. Remember earlier we learned how an automation workflow is like a facto's
64:20 assembly line where each module is like a machine performing a step in the
64:24 process? Well, factories don't operate in isolation. Sometimes your factory
64:28 needs supplies, information, or services from outside its own walls. A real
64:32 factory might call a supplier to order more raw materials. Ask a logistics
64:37 company where a shipment is. Verify parts. Meet quality standards with lab
64:41 testing. request a maintenance crew to check the temperature of a remote
64:44 machine. Simply put, a factory coordinates with external partners to
64:49 handle tasks outside its own expertise. Just like real world suppliers don't
64:52 take factory orders shouted over a fence, external services need a
64:56 structured way to receive requests. That's what an API or application
65:00 programming interface is. In our factory analogy, the API is like the official
65:04 order form that your external partners use to process requests, which you would
65:07 fill out to order some special machine parts to use within your factory. Online
65:12 services like Vappy, Air Table, or Slack all offer APIs that follow the same
65:15 principle. They give outside entities like make.com a structured way to send
65:20 and receive information. APIs give us a clear, consistent way to ask for
65:23 something and get a predictable, reliable response. There are several
65:27 types of HTTP requests that serve different purposes. So, you could be
65:32 saying, "Get me this thing. Post or add this new thing. Put this info where it
65:37 belongs. Replace the whole thing. Patch just this one part of the info, but
65:41 don't replace the whole thing. Delete this thing." In our case, since we're
65:45 about to be making an HTTP get request inside of Make, we're essentially
65:49 saying, "Get me this call summary. Here's what I want, and here's who I am.
65:52 I've got the proper permission to access what I'm requesting." While platforms
65:57 like make.com and n10 provide a visual interface for building workflows, under
66:00 the hood, they're actually making API calls to connect with these external
66:03 tools. So even though you're working with visual blocks, these modules use
66:07 the same underlying language that developers use to connect services
66:11 across the web. Yes, this is no code development, but that doesn't mean code
66:13 isn't running. It's just happening behind the scenes. That was quite the
66:17 detour, but an important one because it gives you a firm grasp on how things
66:20 work on a deeper level. Now that you understand what an HTTP request is and
66:26 how it fits into the big picture, let's set this module up to go get the record
66:30 for the call that was just made. We'll set the method to get. But where are we
66:33 getting something from? Well, we're going to make a request to VP's API to
66:38 get the call that just happened. If we check out the VP documentation, we can
66:42 see that we need to send our request to this URL where the last part of the URL
66:47 is the ID of the call we're fetching. So let's copy that URL and paste it into
66:51 the HTTP module. Adding on the call ID from the vari modules output
66:56 in the header section. Here's where we add instructions like putting a label on
66:59 a package by putting authorization in the name field. This just means we're
67:03 saying we have the authorization to get this info and the proof is the value of
67:07 the header itself. Another way to think of that proof is that we have a key to
67:11 unlock this special box that contains the information we're requesting.
67:17 That key is the API key that we generate from VPY. So in the value field, we'll
67:21 type bearer. Then paste that key next to it. This roughly translates to the
67:26 person who possesses or bears. This key has permission to access what they are
67:30 requesting. Finally, we'll say yes to passing the response. This means the
67:34 module will break apart the response into structured fields that we can
67:38 easily use later in the scenario. Let's make sure this is working by running
67:41 this module only. Here at the bottom, it's asking for the call ID that we want
67:45 to get, which we can just grab off of the last time the VAP module made a
67:50 call. Placing that calls ID here and hitting save will now cause this HTTP
67:54 get request to fire off. And we quickly see that it was a success. We now have
67:57 access to all of this data from our call, including the summary, which as
68:01 you recall is the result of that prompt we added into VP to summarize the call
68:05 for a sales team. We also have access to the analysis, which tells us if the call
68:08 was successful. This is perfect since we'll use both pieces later in the
68:12 scenario to update the lead's air table record with their call summary and
68:15 interest level. So now that we're monitoring the call results, let's start
68:19 implementing those next steps. First, we want to determine whether the call was
68:22 answered. So we'll set up a router to create paths for both answered and
68:26 unanswered calls. When someone answers the call and engages with your voice
68:30 assistant, it's a valuable interaction that your sales team needs to know
68:33 about. Let's set up a filter to check if someone answered the call. On the first
68:37 route, we'll add a filter with a condition that checks for calls where
68:41 the data do the ended reason equals customer ended call. Among all
68:46 the data points available, this is our most reliable indicator that the call
68:50 was answered and didn't go to voicemail. On the second route, we'll do the
68:53 opposite and set the condition to filter for cases where data ended reason does
68:59 not equal customer ended call. On the answered route, we want to know
69:03 if the lead was interested in being sent a proposal or not. So we'll add another
69:06 router that splits into two paths based on the call analysis, specifically the
69:11 success evaluation. Remember earlier when we told Varie to evaluate if the
69:15 call was a success based on whether the lead wanted a proposal or not. That
69:18 evaluation is contained in this data here. Since this data is simply true or
69:22 false, we'll check whether the analysis is equal to true success or false
69:28 failure on each route. For both routes, we want to update the lead's record with
69:31 the call summary and note whether they were interested in a proposal. So on
69:35 each route we'll add an air table update a record module. On the interested route
69:39 we'll target the lead contacts based by this lead's ID and update the contacted
69:45 on field two. Now add the call summary We'll do the same for the not interested
69:58 but check no for interested. Now, our sales rep can instantly see when the
70:01 lead was contacted and what transpired during the call. The system
70:04 automatically documents each interaction, just as a human rep would
70:07 take notes after speaking with the prospect. By the way, in case you're
70:11 curious, this process technically performs a put request under the hood
70:15 where it puts or updates something about the record without replacing it all
70:19 together. Of course, not all of our VP calls will be answered. This happens,
70:23 people are busy or they screen unknown numbers. That's why we build in a
70:26 backup. On the notwans answered route, we'll keep the steps from our initial
70:29 build where we send that lead an email to schedule a call and we let our sales
70:33 team know about them. With this all locked in place, let's run the scenario
70:38 to make sure it's working. Hey, it's Ben from Edge AI. Is now a good time to
70:42 quickly talk about your business needs? Yes, it is. Great. Thank you. I'd love
70:47 to understand a bit more what led you to reach out to us. Are there specific
70:50 processes or tasks you're looking to automate? Yeah, looking to uh implement
70:55 AI agents. That's exciting. Implementing AI agents can really streamline operations. Um are
71:01 there particular tasks or workflows at your company that you find especially
71:04 timeconuming or manual which you're hoping to improve with automation? Yeah,
71:09 just hoping to um have the agents trained on the learning material so that
71:13 they can help my students. That makes a lot of sense. Training AI
71:16 agents to assist your students could really enhance their learning
71:18 experience. Thanks for sharing that. Would it be helpful if I sent a quick
71:23 proposal your way showing how we could support you in developing those AI
71:27 agents? Yeah, that would be perfect. Excellent. I'll send the proposal to the
71:34 email we have on file. Thanks again. We've done a lot. Let's zoom out for a
71:37 second to recap. With our current scenario, we now have a system that
71:41 detects qualified leads, places an AI powered voice call with no manual
71:45 effort, waits, listens, and reacts depending on whether someone answered
71:47 and whether they were interested or not, logging the outcome accordingly. As of
71:51 now, the voice assistant works, but it's not as capable as it could be. Without
71:55 knowing our lead's name or anything about their company and needs, the
71:59 agents ability to pitch is pretty limited. Our next step is to enhance our
72:03 scenario by researching our lead with open AI and passing those insights to
72:07 our Vappy Assistant, allowing it to personalize each call. Essentially,
72:11 we're going to have a a search feature that is going to not only just use the
72:15 voice agent for uh doing the research and getting more information, we're
72:19 going to use OpenAI's uh search models. So, when people fill out that form we
72:23 made, then we're going to use OpenAI to research the internet for that lead and
72:26 get some information on that. Then, we're going to send a call to them again
72:29 using Vappy. This time it's going to be personalized with the information that
72:32 we got from that web search. So it's really, hey, we know this about you, but
72:35 what else? We're looking for this information on you. You can really get a
72:39 a a complete picture of who this person is before we even booked them in for a
72:42 call with our sales team. So with the VIP module we're currently using,
72:46 unfortunately, we're not able to feed anything into it, at least not at the
72:49 time of this recording based on the VIP module's current setup, but we can solve
72:53 for this by switching out the VP module with a more custom approach using an
72:57 HTTP module. So, let's drag it out of the workflow and unlink it. And since
73:00 we'll be borrowing some of the values from it soon inside a new HTTP make a request module.
73:07 Since we're making a manual request out to the Vappy API, we need to specify the
73:12 URL just like we did in our existing HTTP module. So, we can go ahead and
73:18 copy that URL from the get module. Since we're simply placing a Vappy call and
73:22 not retrieving an existing one, we won't need the calls ID. So, we'll leave that
73:26 off. And we'll change the method type to but we need to add a second item where
73:37 the name is content type. Value is application/json. Content type is like labeling the
73:43 envelope you're sending. Application/json means inside this request, the data is structured like a
73:50 JSON object. It's like writing English or Spanish on the outside of a letter so
73:54 the recipient knows what language to expect when reading it. Continuing down
73:58 the module, we'll then set body type to raw, which means we're manually writing
74:02 out the data we want to send to VP. And we set content type to application/json,
74:09 which again tells VPY to expect data formatted as JSON. And I know many of
74:12 you don't know what JSON is, but it's less intimidating than you might think.
74:17 It's really just a pair, the key and the value. Just like in a spreadsheet where
74:21 you have the keys like phones, name, and email. Then you have the values, the
74:26 actual data that goes with each key. It's an easy to read way to structure
74:30 and share data. In JSON, everything sits inside curly brackets. This is called an
74:34 object. Both the keys and values need to be in quotes with commas separating each
74:38 key value pair. Don't worry if you're not a JSON expert. Many free online
74:43 tools can help you check if your JSON formatting is correct. So back in make
74:47 down in the body of our HTTP call out to Vappy, we're going to add some JSON to
74:51 save some time. We'll grab some values from the VPY module we were using
74:54 earlier, including the assistant ID, which we'll paste into the JSON.
74:59 And we'll also grab the VPY phone number ID off that old VP module and paste
75:05 reuse it in our new HTTP module. We're just doing it a bit more manually. The
75:08 main difference here is in this assistant override section. As it
75:12 sounds, we are overriding the assistant within custom variables. These are
75:15 essentially placeholders that will be replaced by the lead's name, company,
75:20 and the research we perform for each of them. This way, when we say, "Hey, Vapy,
75:24 call this lead." We're also saying, "And here's info about them to use on the
75:27 call." Now, that we're going to be sending these variables into the
75:31 assistant, we need to head back over to the Vappy dashboard and tell our
75:34 assistant Ben to be expecting that information and instruct them on how to
75:38 use it. In the first message, we can add the first name variable so our assistant
75:42 can greet our lead by name. In the updated prompt, we'll inform Ben that he
75:46 will receive custom data via variables such as the lead's first name, company,
75:50 and company research. And to use these details to personalize everything on the
75:54 call with the goal of pitching more effectively. With our assistant ready to
75:57 receive all that custom info, we just need to perform research on our lead.
76:01 And for that, we'll be using OpenAI. Let's add an Open AI. Create a chat
76:05 completion module in line just before we make the Vappy call. You'll need to set
76:08 up the connection with your OpenAI account and make sure get an OpenAI API
76:13 key and also have some credits to use for this. To do this, you can log into your
76:19 account at platform.opai.com login and click on API keys. Then create
76:23 a new one, making sure to copy and save it in a safe place. Then you'll click
76:27 over to billing and add some credits, making sure you have an active credit
76:30 card set up. With that connection set up, we'll select the model. Since we're
76:35 doing research, I'm choosing GPT40 mini search preview. By the time you watch
76:38 this, there may be other options. The important thing is to choose a model
76:42 that can do search. In the messages section, we'll set the role to user,
76:46 which just means this message is coming from a user, you in this case. The text
76:50 content is where we place the prompt. Essentially, we're telling it to serve
76:53 as a research assistant that uses our leads information to generate a summary
76:58 of how an AI automation agency could help them with their needs. Let's see
77:01 how it works by running this module only giving it an example company name and
77:11 As you can see, it's running and performing research for us. But notice
77:14 how the result is formatted in paragraphs. While that might look fine,
77:18 it actually causes a problem because when we try passing this result into our
77:22 HTTP module, it needs to follow strict formatting rules. Those paragraph breaks
77:26 can quietly break things behind the scenes and cause the system to reject it
77:30 since it won't be valid JSON. There are a couple ways to fix this. We could
77:34 either demand that chat GPT gives us our summary in JSON, or we could add a text
77:39 passer in between the GPT and HTTP modules, which removes those line breaks
77:42 for us. Since there may be other situations when you need to transform
77:46 data in your future workflows, it's helpful to get some practice passing
77:50 text. So, let's add the text passer replace tool. For the pattern, we'll add
77:53 this regular expression. This essentially finds any line breaks, so we
77:57 can replace them with a new value, which we'll leave blank. In other words,
78:01 instead of the line break, it will be replaced by nothing, aka removed. We
78:05 want global match to be set to yes, so that it finds all line breaks, not just
78:08 the first one, then no, for the rest of the options. Finally, we'll select the
78:12 text we need to pass the result from open AI's research. With this
78:16 reformatting in place, we can now reference that clean text inside the
78:22 company research field of our HTTP body. Finally, we'll say yes to pass the
78:25 response and close out this module. Notice how we still have this red error
78:31 icon on the second HTTP module. It's broken because it's trying to reference
78:35 the call ID from the original VP module that no longer exists because we deleted
78:40 that module. Instead, it needs to reference the call ID from our new HTTP
78:44 module, which manually makes that VP call. Since we haven't yet run that new
78:48 VP replacement module, we'll run it in isolation, feeding it a phone number and
78:53 whatever info just to get it to run. Once it runs, we can now access and
78:57 reference the ID field from it for future runroughs of this workflow. We're
79:02 now ready for the big payoff. Make sure you add a new lead to your Air Table
79:06 with a valid company it can research. Let's run the scenario to make sure
79:15 Hey Mark, it's Ben from Edjai. Is now a good time to talk about your business
79:19 needs? Yeah, sure is. Awesome. I'm following up on the form you submitted. We build custom AI agents
79:26 and automations for companies like View Mastery. Anything from customerf facing
79:30 bots to uh internal workflow optimization. I'd love to understand a
79:34 bit more about what led you to reach out. What sort of processes or tasks are
79:38 you looking to automate? I'm looking to implement AI agents to facilitate the
79:42 learning journey on my e-learning platform. That sounds like a great
79:46 initiative. Enhancing the learning journey for your students can really
79:49 make a difference. Are there any specific bottlenecks or manual tasks
79:54 you're hoping to reduce with these AI agents? Just hoping to increase the
79:57 effectiveness of our content by supplementing it with these AI agents.
80:03 Got it. So, you're looking to enhance the effectiveness of your content by
80:08 providing real-time support to to your students. That's a smart approach,
80:11 especially with the extensive library you have at View Mastery. Thanks for
80:15 sharing that. Would it be helpful if I sent a quick proposal your way showing
80:20 how we could support your goals? Yeah, that would be perfect.
80:23 Perfect. I'll send that to the email we have on file. Thanks again, Mark. All
80:28 right. Thanks. Exciting. Our scenario is working and our voice assistant is now
80:32 empowered to perform dynamic pitches based on our lead's unique information.
80:35 We're almost there. Finally, within build 3, we take this about as far as we
80:38 can within a beginner tutorial like this. Um, and this is a really, really
80:41 powerful thing once you've added in these extra features. So, after the call
80:44 again, we're going to analyze that and deal with if they answered or didn't
80:47 answer. And long story short, if they answered and they said that they were
80:50 interested and said, "Hey, yes, can you please send me a proposal?" Then we're
80:53 going to take all of this research that we've done and all the information that
80:56 we got from the phone call in order to generate them a custom proposal and
81:00 saying hey look this is what we want to kick off with you because you need to
81:02 make a proposal in order to start any kind of services or most kinds of
81:05 services but we can automate the generation of a proposal which can take
81:09 hours and hours and hours for businesses and we can use all of this information
81:12 we collected and they've said yes hey can you send it over and then we use an
81:15 application called Panda do and we can create a template of a document for our
81:19 agency in this case and it's going to use AI in this case we're going to use a
81:24 chatbt uh node here on make and it's going to take all of this information
81:27 and write a personalized proposal on how we would kick things off with them of
81:30 what we're proposing in terms of the scope of work for them like we will do
81:34 this this it sounds like you need this we can do this this is roughly how much
81:37 it's going to cost etc and then using panda do we can send that as a e-ign
81:40 link so that they're ready to sign and we get notifications about if they've
81:43 viewed it if they've signed it etc and so by the time you've done this we've
81:47 automated everything from the initial point of contact where the leader said
81:50 that they're interested in our services to learning more about them to
81:53 determining if they're qualified for our offer uh to sending them a custom
81:57 proposal and ultimately for them signing the dock through Panda do and they're
82:00 ready to kick things off with us. So that is an explanation of of build 3 and
82:04 what we're really trying to go here. I hope this been helpful to clarify things
82:06 for you um because this is really really powerful if you can wrap your head
82:14 Once a lead has expressed interest, it's the perfect moment to harness that
82:17 momentum and transform it into something concrete, a tailored business proposal.
82:21 Why wait for someone on your team to do this manually when we already have all
82:25 the context we need? With the help of OpenAI and Panda do, we can generate,
82:30 send, and log a custom proposal without anyone lifting a finger. So, in the
82:34 final section of this course, we'll be tacking on a proposal generator to the
82:38 end of our workflow. We'll use OpenAI to create the custom text to plug into a
82:42 proposal generator. We're going to be using Panda Do as our tool for creating and sending proposals.
82:48 It allows you to create templates with placeholders that can be filled in
82:51 dynamically from your automation workflows. Here's how to create and
82:54 configure your template in Panda Do. Log into the Panda dashboard. Just create an
82:59 account if you don't yet have one. Then go to templates and click plus template.
83:03 For sake of ease, we can select an existing template to remix such as one
83:07 of these business proposal or advertising sales proposal templates.
83:10 I'm going to use a template I already created here. Within a Panda Do
83:14 template, you are able to drop in tokens, which are basically placeholders
83:18 that can be replaced with actual values, such as your client's company name. In
83:22 this case, you can set up this template however you'd like, but as you can see,
83:26 I've set mine up like this with a client introduction section addressing them by
83:30 name. In the goals and plan section, I've left room to insert information
83:34 about my lead's goals and the services I'll recommend and a plan for how I'll
83:38 implement things. so you can see how it works. We'll create those as variables.
83:43 Over here in the sidebar, we'll add proposal.goals will be a paragraph or two summarizing
83:49 the client's top priorities. Proposal.services will be a bulleted list of the
83:55 recommended services. Proposal.implementation will be a concise execution plan to
84:01 deliver the above. In the pricing section, I've already added placeholders
84:05 here. proposal pricing and for a breakdown of services and costs and
84:08 proposal.total, the full estimated project total. Down in the agreement section, we're
84:15 requesting signature and including the leads info. Once everything is in place,
84:19 name it something relevant and save it because you'll soon be using this
84:23 template inside your makes panda dooc module. Optionally, you could spend some
84:27 time styling this template with a logo, brand colors, etc. Remember, this
84:31 document will be client-f facing, so make it look and feel as professional as
84:34 the service you're offering. With your Panda do template ready and tokenized
84:38 correctly with placeholders, we're ready to include it in the final sequence of
84:42 our make scenario. We'll add the Panda Do create a document module, set up a
84:46 connection with our Panda Do account, name the document based on the company
84:49 we're sending it to, and select the proposal template we created earlier.
84:53 We'll fill in these values, giving the module the lead's email to send this to
84:59 and include all of the necessary info about our client, like the company name
85:02 and the client's first and last name. For all these proposal tokens, we'll be
85:06 generating these values with AI in a moment. For now, we'll scroll down and say yes
85:12 to send a document because we want to email this Panda doc to our lead. Fill
85:15 in the subject line using their name and company, write a short message,
85:22 then hit save. Next, we'll use OpenAI, create a chat completion module to help
85:26 us write a clear, convincing proposal. We'll select a quick and efficient GPT
85:30 model, and in the prompt, we feed it instructions about its role as a sales
85:35 expert with context about the services our mock company offers like AI
85:39 automation and agent-based systems, information about the client, including
85:43 the company research we did earlier in the workflow, and the summary of the
85:46 call our voice assistant had with them. Then we clarify its task to identify the
85:50 most relevant services we can offer them and we demand the output to be in JSON
85:54 format so we can make easy use of it in the modules after this.
85:59 If we go ahead and run this module only passing in some dummy data for the call
86:10 We'll see it efficiently generates this relevant info organized as JSON like we
86:14 requested. We're almost ready to plug these values into the Panda doc module,
86:18 but we first need to add a pass JSON module to prep the message content
86:22 output from OpenAI, breaking it up into discrete values that are available as
86:25 variables we can inject into our proposal document in the next step. To
86:29 plug these proposal values into their slots in the Panda do module, we first
86:34 need the JSON module to run. While we could run our entire scenario, there's a
86:38 quicker way. go into our OpenAI module, grab the expected JSON format, and paste
86:44 it as dummy data into the JSON module to test it. Perfect. Now, we can see it's
86:48 passing correctly, and we can now go into our Panda Do module and plug those
86:52 variables into their corresponding token Finally, we want to bring things full
87:00 circle by updating the leads air table record so that our sales team can see
87:11 So in the proposal sent on field, we'll add a timestamp with now. This keeps
87:15 everything tidy. So when someone on your team looks at a lead, they know a lead
87:19 was created, they were qualified, they were called, including a call summary,
87:22 they were interested, a proposal was sent. No one has to chase a status
87:26 update. It's all there in one glance automatically. Let's do a final run of
87:30 this scenario to make sure all of the moving pieces are working correctly.
87:37 Hey Brad, it's Ben from Edge AI. Is now a good time to quickly talk about your
87:41 business needs? Yeah, it is. Great. Thanks. I came across some info about
87:45 View Mastery. It sounds like you're really focused on helping developers
87:49 learn VDJs efficiently. We build custom AI agents and automations for companies
87:53 like like yours. Anything from customerf facing bots to internal workflow
87:57 optimization. I'd love to understand a bit more about what led you to reach
88:01 out. Are there specific processes or tasks you're looking to automate? Yeah,
88:05 I'm looking to implement AI agents for learning who are trained on the content
88:10 library of my learning platform. Uh, that makes a lot of sense. Creating
88:14 AI agents that can assist learners by leveraging your content library sounds
88:18 like a great way to enhance engagement. Are there particular bottlenecks or
88:21 manual tasks you're hoping to reduce with these AI agents? Yeah, just hoping
88:25 for uh the students to be able to comprehend and apply the concepts more
88:30 effectively with the agents. That's a smart approach. Empowering
88:37 students to better understand and apply concepts with AI support can really
88:40 boost their learning experience. Thanks for sharing that. Would it be helpful if
88:44 I sent a quick proposal your way showing how we could support new mastery with
88:49 tailored AI solutions? Yes, please go ahead and do that. Perfect.
88:54 I'll send the proposal to the email we have on file. Thanks again, Brad. Thank
89:00 you. You're welcome, Brad. Thanks for your time today. If you have any
89:03 questions later or want to discuss further, just reach out. Have a great
89:08 day. The workflow seemed to have worked as planned. Inside Air Table, we can see
89:13 that the proposal sent on date was added. Now, let's check the email that
89:16 we told Panda do to send the proposal to. Heading into that email, we can see
89:20 that it worked perfectly. Our proposal dock arrived which the client can open
89:24 up to go view their proposal which we can verify has been properly drafted
89:28 detailing the client's specific needs and our plans to help them. From here
89:32 the client can sign the document, date it and finish it out.
89:36 An added benefit of Pandanda is that we don't need to create additional
89:40 automation steps in make.com to track when leads view or sign proposals. Panda
89:44 handles this automatically by sending notification emails to our email
89:48 whenever a lead views or completes a proposal. With this final stretch,
89:52 you've created a full system that not only identifies and qualifies leads, it
89:56 translates interest into action. Of course, there are many other steps you
89:59 could add to this scenario. For example, the delay we added after the call takes
90:03 place works in most cases, but if the call exceeds 5 minutes, the system would
90:08 incorrectly mark it as not answered, even if it was successful. The foolproof
90:12 solution would be to implement a web hook that listens for the call to end,
90:16 but that's outside the scope of this beginner build. You could even add a
90:19 whole extension to this workflow where you detect when a lead signs the
90:23 proposal and then onboard them with an automated client orientation workflow. I
90:27 encourage you to get creative and add on to it to learn and challenge yourself.
90:30 For now, I want to share some troubleshooting tips to keep in mind as
90:33 you go off on your own and build your own workflows. [Music]
90:39 As you take your next steps and start building your own AI powered
90:43 automations, I want to be completely transparent. You will encounter issues.
90:48 platforms will change, tutorials may become outdated, and unexpected problems
90:52 arise. This isn't a flaw in the system. It's a natural part of working in a
90:56 rapidly evolving field. The truth is, even experienced developers spend a
91:00 significant portion of their time troubleshooting. I can't count how many
91:03 times I found myself yelling at my screen because some seemingly simple
91:07 thing wouldn't work. It's part of the process for everyone. There's an old
91:11 saying, give a man a fish and you feed him for a day. Teach a man to fish and
91:17 you feed him for a lifetime. This sums up our approach to technical education.
91:21 If you are spoonfed every solution, you'll end up just copying existing
91:25 systems and won't be prepared for real world challenges and you won't develop
91:28 ways to differentiate yourself either. So, think of technical problem solving
91:32 as a muscle. Right now, it might be underdeveloped and using it feels
91:36 uncomfortable. That's normal. But with consistent exercise, tackling problems,
91:40 finding solutions, learning from mistakes, this muscle will grow
91:44 stronger. It's a gradual process that develops through experience. Each
91:48 problem you solve builds your expertise, making future challenges less daunting.
91:52 While you can't rush this journey, you can fully embrace it. When you feel that
91:56 sense of being out of your depth, try to recognize it not as failure, but as
91:59 growth. You're in uncharted waters. You're pushing your limits and expanding
92:03 your capabilities. The most valuable learning happens precisely in these
92:07 moments of struggle. So with that growth mindset, let's cover some tools you can
92:12 use for troubleshooting. You probably won't be surprised that I encourage you
92:15 to leverage AI as you develop your skills. If you run into errors in your
92:19 workflow by heading over to chat GPT or a similar LLM, you can lean on its
92:23 ability to support you in solving the issues you encounter. The trick here is
92:27 to describe your issue in detail. Remember to give enough context so the
92:30 AI can grasp what you're building, the specific step you're on, and the
92:34 expected outcome versus the problem you're experiencing. Include
92:38 screenshots, error codes, and let it There will be times your issues are a
92:45 little more complicated. For complex problems, you can even try the deep
92:48 research mode to get comprehensive information on specific platforms or
92:52 issues. Sometimes the first suggested solution won't be the right one. But as
92:56 you engage in back and forth dialogue, as you see me doing here, where I am
93:03 the AI gains more clarity about your issue and can better narrow down
93:11 And if you truly want to learn, don't just do what it says without curiosity.
93:14 Ask it to help you understand why the issue happened and why the new solution
93:18 ultimately worked. If you are hitting a wall, it's possible that others have
93:22 already hit that same wall and climbed over it. So, by doing a traditional
93:25 Google search, you can look for existing threads or community posts about similar
93:30 issues. Many problems you'll encounter have already been solved by others. Of
93:33 course, YouTube is another great resource. Search for recent tutorials
93:37 that show updated interfaces and workflows or that highlight key details
93:42 that could help resolve your issue. We now have access to tools like Google AI
93:46 Studio that can watch your screen and provide real-time support as you work.
93:50 Think of this as an experienced guide peering over your shoulder and helping
93:54 you step by step. This eliminates the need to manually type out your problem
93:58 and provide screenshots to chat GPT since you can give the tool permission
94:01 to observe what's on your screen and guides you through solutions using voice
94:05 assistance. Take a look at how powerful this is as it helps diagnose what's
94:09 causing this error. I'm getting an error on this HTTP module where I'm trying to
94:13 make a post request out to the VP API. Any idea where I should start looking
94:18 for a solution to this error? I see the workflow you've created and that you are
94:21 receiving an error when trying to make a post request to the VAPI API. Are there
94:28 any error messages that you can share? It's a 400 bad request error. Do you see
94:34 the information here in red? Okay, I see that it's a 400 bad request error. The
94:39 message indicates there's an issue with the JSON formatting of your request
94:42 body. It seems like there's an unexpected character, either an expected
94:47 or after array element in JSON at position 1167, line 12, column 51. Would
94:53 you like me to help you review the JSON structure in your request to identify
94:58 any syntax errors or missing commas or brackets? Yes. Can you notice anything in here
95:03 that is wrong? Looking at the JSON, it seems like the error might be in the company research
95:09 field where you have an array containing a single element and that element seems
95:13 to be an expression. If you're trying to pass a single string there, try removing
95:17 the curly bracket. Also, can you double check that all your keys and values are
95:21 properly formatted as strings? Learning to read and apply documentation is a
95:26 critical skill. Each platform has detailed guides that are regularly
95:29 updated. Often, you'll find information there that can help solve your issue. If
95:33 it feels overly technical, you can circle back to asking something like
95:36 chat GPT to help you understand the documentation and how it applies to your
95:41 specific issue. Finally, there are plenty of online communities like the
95:45 ones me and my team run related to the platforms you're using. Join Discord
95:48 servers or other forums, search for similar issues, or even ask questions
95:52 directly. You'll be surprised by how helpful and supportive these communities
95:56 can be. If your goal is to build real automations for real clients, things
95:59 won't always go according to plan. platforms will change and requirements
96:03 will shift. The ability to adapt, troubleshoot, and find solutions is what
96:07 separates successful builders from those who give up early. The most successful
96:11 people in this field aren't those who never encounter problems. They're the
96:14 ones who persist through difficulties, who see obstacles as puzzles to be
96:18 solved rather than roadblocks to progress. So, when you hit that
96:21 inevitable moment where something isn't working, and you feel stuck, remember
96:26 this is normal. Take a breath. Step away if needed. Then come back and work
96:29 through your troubleshooting toolkit. Each time you solve a problem, you're
96:33 not just fixing that specific issue. You're becoming better at solving all
96:37 future problems. This resilience and problem solving ability will be just as
96:41 valuable to your success as any technical skill you learn. So, embrace
96:44 the journey of building your technical skills. In the next and final section of
96:49 this course, I'll show you how to sell your newly built systems to real world
96:56 Now that you understand how AI automations work and can build them for
96:59 yourself, let's talk about actually making money with these skills. But
97:03 first, let me destroy a huge misconception that many people have. You
97:06 don't need to build the next chat GBT or create some revolutionary AI startup to
97:10 make some money in the AI space. The real opportunity is much simpler.
97:14 Helping businesses to understand and implement AI automations like what
97:17 you've just learned. This is how I monetize my AI automation skills, and it
97:21 has been the most explosive growth I've ever experienced in my career. The good
97:24 news is, if you've made it this far in the video, you're much closer to being
97:27 able to tap into this starving market for AI automation services than you
97:31 think. But don't take my word for it. Let's hear from some of the world's most
97:34 famous and successful businessmen. If I was 25 years old today in 2024, what
97:39 would I do? What's a good sector to get involved in? What business would I get
97:42 involved in? I think everything is looking at AI now in a different way.
97:46 And I think AI growth is going to be exponential. So, anything to do with AI.
97:50 Now, what could that be? in the simplest form is helping people use the
97:54 technology. There's going to be a massive amount of people wanting to use
97:57 it that don't know how to and they're willing to pay to solve that painoint.
98:03 So, is that consulting? Not really. It's implementation and execution and so
98:08 helping a business do that transfer into a world where they're controlling their
98:12 data and getting information from it. Now, the majority of businesses in
98:16 America, for example, are between 5 and 500 employees. So, they're small
98:21 businesses. They create 62% of the jobs. They want to use AI. You should help
98:25 them solve for that and they'll pay you. Mark Cuban is saying the same thing as
98:29 well. That the biggest opportunity right now is helping small to mediumsiz
98:32 businesses who don't understand AI and automation yet but desperately need it
98:36 to survive. And they're right. Here's why. According to recent data, there are
98:40 1.7 million businesses in the US alone, making between $500,000 and $10 million
98:45 per year. These are small businesses which make up 62% of all jobs in the
98:49 USA. These businesses know that they need AI automation to stay competitive,
98:53 but they don't have the time to learn it themselves. And there's basically no one
98:57 there to help them at all. All of the big consulting firms are looking at
99:01 other big businesses, and no one serving this small business market who still
99:04 need AI automations just as much as anyone else. So basically, almost all
99:08 small businesses are starving for AI automation services like education
99:11 services in order to help them to understand what AI and automation is and
99:16 why they need them. Then there's consulting services which help them to
99:20 identify where AI automation can help them the most. And finally, there is
99:23 actual implementation services to help them build and maintain their AI
99:27 automation systems. Right now, based on data I've collected in my community, for
99:30 every person offering AI automation services, there are over 1,100
99:35 businesses in the USA alone that need help. 1 to 1,100. The market is
99:39 completely untapped and will be for a very long time. And that's where you
99:42 come in, which is helping these hardworking small business owners to
99:46 understand and implement AI automations so that they can keep up and survive
99:49 through this AI revolution just like you. We've seen this exact same pattern
99:53 happen when the internet came out. The companies that help businesses to adapt
99:56 to the web made fortunes as the companies that they helped. I personally
100:00 spotted this opportunity in early 2023 and started Morningside AI, my AI
100:03 automation agency. And since then, I've generated over $5 million selling AI
100:07 products and services. And as a company, we're only just getting started. And the
100:11 best part is that, as we've proved in this video already, that you don't need
100:14 to be a technical genius to understand AI and even to build your own AI
100:18 automations. You just need to be one step ahead of the businesses that you're
100:21 going to help. Let me show you the three specific ways that you can start making
100:27 money with your AI automation skills. So, there are basically three services
100:29 that you can provide to help businesses with AI automation. First, there's
100:34 education. This is teaching businesses about AI automation, running your own
100:38 workshops or presentations, or training their teams and creating courses.
100:42 Businesses are desperate for someone who can explain this stuff in simple terms
100:45 to you, just like I've done for you in this video, what the hell AI is, what
100:49 automation is, and what it can do for them. And after watching this video and
100:52 a few others that I've made on this channel, you'll know more than enough to
100:55 start educating businesses and helping them to sort of move from where you were
100:58 at the start of this video to where you're going to be at the end of it.
101:00 I'll be covering which videos those are in a moment. And secondly, you have
101:03 consulting services. And this is where you analyze a business's operations and
101:07 show them where AI automations can help them save time or make extra money.
101:11 You're essentially being their AI automation strategist. For example, you
101:14 could recommend a lead qualification workflow like the one you just made to
101:19 help a struggling sales department. And third is implementation. This is where
101:22 you actually build and deploy AI automation solutions for businesses. Or
101:26 better yet, like my agency, you can do all three of these. You can do
101:28 education, consulting, and implementation. But you don't have to do
101:32 it all at once. It took us like over two years to get here. So there's no rush.
101:35 And believe it or not, there are people with only a few months of experience in
101:39 the AI automation space selling all of these kinds of services right now.
101:41 Education, consulting, and implementation. And the demand from
101:45 businesses is increasing insanely fast right now because they're all switching
101:47 on and realizing they need to do something about this right now. But
101:51 here's the thing. You still have one problem, which is that you still don't
101:55 really know enough. You're close, but you're not quite there. The way to make
101:59 money in the AI automation space or with any services really is to create what's
102:03 called a knowledge gap between yourself and the people that you're trying to
102:07 help. Your knowledge gap is your money maker. Businesses will pay you in
102:10 proportion to how much more you know about AI automations workflows and their
102:14 business applications than they do. So now while this video has taught you a
102:18 lot, your knowledge gap is still fairly small. But we can fix that. Let me break
102:21 down exactly what you need to do to extend your knowledge gap so that you
102:27 can start monetizing this skill. This video is step one. As long as
102:30 you've been taking notes and followed all of the tutorials by building the
102:34 automations alongside me, you're already far ahead of most people who have no
102:37 idea about how to build AI automations. Step two is building even more
102:41 experience with AI automation so that you are more familiar with the platforms
102:44 and better understand different ways that they are being used to help
102:47 businesses. I've only given you a taster here. This is a foundational knowledge
102:50 is the point of this video. But to do this and extend your knowledge gap, you
102:54 can take my free course on school where you'll build another 5 to 10
102:57 automations. Um, and the link to join will be in the description below this
103:00 video. And this will further expand your knowledge gap without paying a dime.
103:03 Once you've done that, you'll have what I call a foundational knowledge. So you
103:07 understand the core AI concepts. You can build basic solutions yourself. And you
103:10 know what's possible for businesses right now. And then comes the big
103:14 decision. Do you want to go deeper technically or do you want to start
103:17 monetizing what you already know? As we've already covered, building and
103:20 implementing AI automations is only one of the services that you can sell.
103:24 Naturally, the technical skills needed to make money in implementation,
103:29 actually building AI automation systems are far greater than just having a
103:33 foundation, you know, so you need to be a lot more skilled in terms of
103:36 experience and being hands-on with the technology to deliver high-quality
103:39 services for your clients. However, with a good foundation, you're basically
103:42 ready to start having a crack at selling AI automation, education, or consulting.
103:46 So this decision of whether you go to education and consulting or actually
103:50 building it depends on what you're really interested in. I'll use myself as
103:54 an example. I've always loved making things from like block houses when I was
103:58 a kid to like brewing beer with my my granddad to tinkering with engines. So
104:02 when I hit this foundational level in early 2023 with my skills, I naturally
104:05 kind of just dove deeper into the technical side. I kept building more and
104:10 more complex AI automations which led me to starting Morningside AI where we now
104:14 build AI solutions for clients. But here's the thing. A lot of people aren't
104:17 like me. And chances are you aren't either. They don't get much of a buzz
104:21 out of the building side of things. And because of this, many of you going to be
104:24 better at teaching and actually working with people to help them understand and
104:27 get value out of this technology than actually building stuff. But that
104:30 doesn't mean that you can't make money in the space for you. Using the
104:33 foundational knowledge that you'll have after you finish that free course to
104:36 actually sell AI, automation, educational consulting makes way more
104:41 sense. So the key is being honest about your strengths and your interest and
104:44 setting real expectations on how technical you want to get in your career
104:47 by picking whether you want to sell educational consulting or development
104:51 services. You have a hard stop on how much you need to learn before you start
104:55 taking action. So this essentially prevents you from getting stuck in an
104:57 endless learning phase. I see it all the time where people just keep, oh, I don't
105:00 know enough. I need to keep learning and they're forcing themselves to do
105:03 something they're not really good at or they don't want to do and they're
105:05 procrastinating when they could be out there making money. So, in summary, your
105:09 options from here are if you love building and want to learn more, then
105:12 just keep going. Keep following that interest and that energy. Watch my free
105:15 course tutorials in the description on my free school. And then after you've
105:18 completed all of those, go on to build your own projects and then ones for
105:21 friends and family. Try solving your own problem or the people around you solving
105:24 their problems with AI workflow automation. And within 2 to 3 months of
105:27 doing that, you'll be ready to start trying to sell implementation to the
105:30 people around you properly. However, if by now you haven't fallen in love with
105:34 the building process, then it's probably best that you just finish off your
105:36 foundation by doing the builds in my free course and then get started on
105:40 monetizing. Like, you don't need to keep forcing yourself to learn more and more
105:42 and more and more. You already know enough cuz you have a knowledge gap
105:45 between yourself and your clients and you can start to help them by monetizing
105:50 that knowledge gap. So, once you're clear on what type of AI
105:53 automation business owner you are going to be, getting your first few clients is
105:56 actually pretty straightforward. And there are two main ways of doing this.
105:59 The first is through what are called warm connections. And this is by far the
106:02 easiest way to start. Instead of expecting strangers to trust you with
106:06 your automation expertise, you can start with people who already know you and
106:08 already trust you. What you're going to be doing is basically making a big list
106:11 of all the people you know, all of the connections and friends that you have
106:15 and acquaintances even, and reaching out to them systematically to say, "Hey, I'm
106:18 doing this AI automation thing. Would you be interested in having a chat to
106:21 see how we could help you or your business?" And it's just consistently
106:23 starting those conversations with these people in your network. And then
106:26 eventually, one of those doors is going to open and become your first client.
106:29 I've covered this exact process many times here on the channel. So on the
106:31 school post for this video, I'm going to add the all of my complete guides for
106:35 warm outreach, including resources directly from my AAA accelerator program
106:39 that will help you to fast track your first few clients. The second way is
106:41 what I call the community content flywheel. So this is how you build
106:45 long-term momentum with getting clients for your business. Firstly, you need to
106:47 get inside my free community on school and then immediately start creating
106:51 content about what you're learning. So this could be uh making YouTube
106:54 tutorials about building automations. It can be LinkedIn posts about workflow
106:58 automation tips and tricks or whatever platform you prefer. But here's the key
107:01 is that you need to share this content back into the community. And we have
107:06 over 160,000 members in my free school with the biggest AI business community
107:09 in the world. And that gives you an instant audience of people who are
107:12 interested in the same things to help you get traction with your content as
107:15 fast as possible. So let me give you a perfect example of this working in
107:18 action. We have a guy called Rory Ridges, a young guy from the UK, and he
107:21 joined my free community and followed this exact process. He took the free
107:24 course in the community. He learned the basics and started posting simple
107:29 tutorials on make.com and relevance AI and literally just started sharing the
107:32 automation workflows that he had learned from my videos. Like he wasn't trying to
107:35 reinvent the wheel. He was just saying, "Oh, I learned this cool thing. I'm
107:38 going to make a video on it." But every time that he made a tutorial or a video,
107:41 he would share it back into the community. The community would watch it,
107:44 give him feedback, and many of them would go and subscribe to his channel
107:47 and become regular viewers. So, this not only helped him to grow his channel
107:50 faster, but it started to position him as an automation expert to his potential
107:53 clients. So, now his YouTube channel brings him in enough leads to support
107:57 his growing AI automation agency. He's basically started the same flywheel that
108:01 took me from zero subscribers and 0 with AI to on track to making $10 million
108:06 this year and over 500,000 subscribers in just 2 years. So, essentially, the
108:09 community gives you an audience. The content gives you credibility and
108:12 together they bring you clients consistently. So, on my free school,
108:14 there's going to be a post for this video, and I'm going to leave all of the
108:17 links to my complete guide for creating content and generating leads, just like
108:20 Rory and I have done. I've done a video on it, ton of resources. You can find
108:25 that all on the free school. What's really important to notice with
108:28 both of these methods is that they start with giving value first. Whether it's
108:32 helping your warm connections to understand and implement AI automation
108:36 for free or sharing your workflow automation knowledge for free through
108:39 content, you have to start giving before you get. Now, I know all this businessy
108:42 stuff may feel a little bit overwhelming or out of reach for some of you, but you
108:47 will seriously be amazed at what baby steps add up to in this AI automation
108:51 space. You've already taken the first step by watching and following through
108:55 on this video. So, congratulations. All you need to do now is keep this momentum
108:58 going. And the next step for all of you is pretty clear now. You need to jump in
109:01 my free community. Go in there and drop an introduction post. Let everyone know
109:04 who you are, what you'd like to do with your AI automation career, and then
109:08 start working through my free course material. It's there for a reason. And
109:11 I've poured everything that I've learned about AI automations and building AI
109:15 automation businesses into videos like these and they're all in a nice sequence
109:18 for you to work through on school each time you complete a video. You can click
109:21 the little check box and keep stacking those small wins and keep that momentum
109:24 going until you get to where you want to go. And all of the resources I've
109:27 mentioned in this last chapter, the selling part of this video will be on a
109:31 school post for this video within school if you go to the YouTube resources tab
109:34 and then it should be right there. And don't forget to check all of those
109:37 resources out there. And of course, if you've made it this far, could you
109:40 please do me a big favor and just leave a like on this video? You can drop a
109:43 comment below. Tell me what you like the most or what you'd like to see next. And
109:46 click the share button and send it to any loved ones or family or friends or
109:49 anyone so that they can start learning these valuable automation skills for the
109:53 future as well. So, all of these actions just help my videos like these reach
109:56 more people in the YouTube algorithm and I'd really appreciate it cuz I put a lot
109:59 of work into these for you. And of course, subscribe to the channel for
110:02 more content like this, helping you understand AI automations, how to build
110:05 them, and more importantly, how to build businesses and make money around this
110:09 incredible opportunity that is AI. If you want to check out my 4-hour guide on
110:12 how to build AI agents, I've got that up here. It's just like this one, but on AI
110:15 agents. Got a ton of great feedback from it. So, if you want to keep learning,
110:18 that's a great place to go. But aside from that, guys, thank you so much for
110:20 watching. That's all for the video and
