0:02 I was finally able to build an AI automation to handle a project for me
0:05 that I've been trying to build for a few months now and I used Make AI agent for
0:10 it. So, I partnered with Make for this video to show you exactly how it works.
0:13 Now, I'll give you a quick background on this and I think a lot of people could
0:17 benefit from this type of automation. I've been using ton of different tools
0:21 in the last couple years like custom GPTs inside of chat GPT to repurpose my
0:26 YouTube videos and it's never really clicked for me because it still requires
0:31 a lot of manual steps in the automation process, right? I still have to click
0:35 things, copy and paste things and it's something that I continuously keep
0:40 forgetting to do in my day-to-day work. Now, with MakeAI agents, I was able to
0:45 build a real AI automation that handles every step for me. So, if you haven't
0:49 used Make before or if you're new to it, Make is a visual platform where you
0:54 could design, build, and automate just about anything. Now, the new upgrade,
0:58 Make AI agents, they let you build customizable AIdriven assistants that
1:04 could decide, adapt in real time, and integrate with your workflows, too. And
1:08 make could connect just about anything. Over 2,000 different apps across 30,000
1:12 different actions. So, pretty much whatever you want to build, this could
1:15 probably connect it. And I've set myself a brand new account here so you could
1:19 follow along. Any type of prompt that I'm going to use. I'm going to include
1:22 in the description. And before we jump in, make also gave me a free month of
1:27 the pro plan with 10,000 operations included that I could give to you. So
1:31 I'll put a link in the description to that plan so you could follow along with
1:35 the rest of the video. And I think this is something you'll want to build for
1:38 yourself, too. So here inside of my make account on the left side, you'll see
1:43 something called AI agents. So click on this and we'll go ahead and click create
1:47 an agent to create our very first one. Now the first thing it's going to ask
1:50 you is going to ask you to create a connection. Now this connection is
1:54 simply asking what AI model you want to use. That's the nice thing with this. It
1:58 gives you ton of different option. Now I'm going to choose open AI. Make has
2:01 their own AI provider. You could use Anthropic Gemini. Pretty much any AI
2:06 tool you want is under this dropdown. But OpenAI the creators of chat GPT
2:09 because I'm going to use a chat GPT model. I'm going to choose that one.
2:13 Then it's going to ask for my API key here that comes directly from OpenAI. So
2:18 if you're totally new to these type of AI automations, every time you want to
2:22 create some kind of automation, you will need to get an API key from the company
2:27 that runs the AI model. In this case, I'm going to use a version of chat GPT.
2:32 So that comes from OpenAI. Let me show you exactly how to get that right now.
2:36 If you don't know how to get an API key from OpenAI, it's available to everyone.
2:40 So, if you go to a website called platform.openai.com, it's going to bring you to this page and
2:46 you want to go ahead and log in or create yourself a free account. Now,
2:50 this is totally separate from your chatgpt.com account. This is to get an
2:55 API key so we can build those kind of automations. I'm going to go to the
2:58 dashboard here and on the left side, you'll see API keys over here. Now, I'm
3:02 going to go ahead and create a new secret key. You could name it here and
3:06 create it. And most likely if it's your first time creating one, it's going to
3:10 ask you for a credit card and you'll pay open AAI on the OpenAI platform based on
3:15 API usage. And back on make it will ask you to go to the OpenAI billing page
3:18 here. So it will link directly there too. I'll save it now. Next, I'm going
3:23 to name this agent script repurposing agent. And then I'm going to choose an
3:27 AI model. I actually like chat GPT 5 mini. Now, this is a good balance in my
3:32 opinion between a really smart AI model, a fast AI model, and a very affordable
3:37 AI model from all the ones available. Then you have your system prompt right
3:42 over here. So, system prompt is what the agent needs to do the task that you're
3:46 going to give it. You want to describe exactly what you want this agent to do.
3:50 Now, I'm going to paste mine here. Then I'll walk you kind of briefly through
3:53 it, but I'll copy it in the description below the video so you could kind of
3:57 copy and paste it for your own and go ahead and edit it as you need to. Okay,
4:01 it brings us to this page here and I'll walk you through the different settings.
4:04 This is where you set up your MakeAI agent. Here you could click on improve
4:08 which I found really beneficial. It improves your system prompt especially
4:12 if you're new to system prompting. But to walk you through the basics of the
4:17 system prompt, I tell it what it is. So you are a script repurposing agent. goal
4:22 is to repurpose YouTube scripts that I write in Google Docs into two new
4:27 formats. So, the point of my agent, and you could do this with a blog post, for
4:31 example, any type of document that you have inside of Google Docs. I actually
4:35 don't have these videos on YouTube. I'm not pulling them from YouTube. I pull
4:39 them directly after I write the script in Google Docs. I'll show you how this
4:44 all works. Then what it does is first it creates a 60-second YouTube short script
4:49 with a strong hook, three quick steps, and one clear CTA. So it writes a new
4:54 script for me that's ready to read. I'm going to show you what it does after
4:57 that, too. It writes a LinkedIn post for me, and then I told it exactly what to
5:02 do with those things. I told it when it's done writing that YouTube short
5:07 script to ping me on Slack to tell me it's ready. So I go ahead and record it
5:11 like this. I like to create original YouTube shorts and then on top of that
5:16 it will also post the LinkedIn post for me. It will write one create a document
5:21 around it and post it for me and I could build on top of it even more. But right
5:25 now that is doing two things that I was always leaving on the table that I just
5:29 didn't get the time to do. They were not fully automated that required copy and
5:32 pasting. This does it for me automatically. Then you have context. So
5:37 this section is also a very important section. background information,
5:42 additional context for your agent. So, what I like to do here is actually tell
5:46 it the tone and the style of the writing, but I want to show you how to
5:50 create a context document from scratch in this step because this is something
5:54 you could use across all your AI automations. So, because this is a
5:59 repurposing agent, I needed to have a really clear understanding on how I like
6:03 to talk, my tone, my reading level for the content that I write. So, what I'm
6:07 going to do is inside of chat GPT here, I'm going to give it a simple prompt
6:10 like this. Analyze the following text for writing style, the tone, the reading
6:15 level, and create a document for me using canvas. This is a tool that chat
6:19 GPT has just to make it easier to read and download that I can give an AI agent
6:24 to mimic my writing style. And then I'm just going to paste something that I got
6:29 from a YouTube transcript. So, I just went to YouTube on one of my videos and
6:32 I copied the transcript. You could do this with anyone's writing or your own.
6:36 And I recommend if you wrote anything, as long as there is enough like a
6:39 paragraph or two, it will have enough of a sample here to analyze. And you can
6:44 see writing style, conversational, and really, really detailed breakdown, tone,
6:49 reading level, whatever else you want to analyze with your writing style, I
6:52 recommend you do that here. Okay. Now, when it's done, I'm just going to press
6:55 download right on top. And you could download any of these formats. Now,
7:00 markdowns are fine, too, but usually a Word doc is fine. Now, back on make, I'm
7:03 going to add this context. And again, you could do PDFs, text files, any type
7:08 of file. Just analyze your own writing style. I'm going to upload mine. And
7:11 then I'm going to add a description because you typically want to have a
7:14 really solid description with pretty much every step of building an AI gen so
7:18 the agent knows everything that is going on with this. So the context of this is
7:23 this is my writing style. Use it anytime you repurpose any content. So I'm going
7:27 to go ahead and save that. Now you have two more things here as you build your
7:31 agent. So, so far we had our AI model that we picked in the very beginning. We
7:35 have the system prompt. Those are the two most important things to get you
7:39 going. Then the context is optional, but I like to give it something like I just
7:43 showed you. Then you have tools. So, what makes an agent really powerful is
7:47 access to other tools. And there's a couple of things you'll see over here.
7:51 First, you have MCPs. So, MCPS are a connector. They could connect to pretty
7:56 much any type of an external server. So that's one way to connect to just about
8:01 everything on the internet. But in this case, we're going to use tools, which
8:05 are also things that could connect make to other services. So with tools, you
8:09 have two options. I'm going to press add here. You could add to modules or
8:14 scenarios. Now, this is really interesting because modules could be
8:18 simple individual actions. And I have an example for you to see, but if you want
8:22 one action to be done, a module is going to be able to do that. connect to one
8:26 tool and one action. Scenarios are entire workflows that you could connect
8:30 to and you could give the agent the power of that entire scenario, which is
8:35 a really powerful option. And you could pick and choose multiple module
8:40 scenarios this agent can have access to. And once the agent does have that
8:43 access, it will be able to make those decisions and complete those tasks for
8:48 you using those tools. Okay, let's go ahead and start here with a couple of
8:51 tools that we need. The first one I'm going to create is going to be a
8:53 scenario. So, it's going to be a multi-step tool. Now, you could go ahead
8:57 and create it here, right? Since we don't have one, it lets you create it
9:00 right over here. I typically actually like to go to that page, scenarios page
9:06 right here. And you'll see all the different scenarios you've built in the
9:10 past because you could use make just with scenarios without the AI agents.
9:15 Obviously, when you give access to the scenarios to the agent, it gets a lot
9:18 more powerful and it could start making decision and do things a lot more than
9:23 just using the regular scenarios here. But I'm going to go ahead and create one
9:26 here right on top and I'm going to go ahead and build one from scratch. They
9:30 also have ton of different templates, but in this case, I know exactly what I
9:33 want to happen. This is what the page looks like. They got ton of different
9:36 tutorials if you haven't built these before. Now the very first thing I like
9:40 to do is I just like to come up here and I like to name the scenario so I don't
9:44 forget a lot of times if I start creating it I forget to name it later
9:48 and with the agent the instruction you give it a lot of times you could call
9:52 this by name this type of a tool or a scenario it will go and fetch that and
9:56 then before I press the plus sign with each of these down here you have an
10:01 input and an output I like to set this right up front also and then I'll show
10:05 you exactly what's going to happen when I press the plus sign But with the
10:09 input, let me add one here to explain what's happening. I'm basically going to
10:13 tell it what I'm expecting to happen as the input and what the output should be.
10:17 Now, I call a script text from a file ID. You could name this whatever you
10:21 want, but it's best to have some kind of a naming system that is easy for you to
10:25 remember. So, I have one here that I've created. Then, I describe what's going
10:29 to happen. File ID from a Skillap YouTube video, and it will get the text
10:33 document from that. It's in text format. I leave default value to empty. And I'll
10:38 set this to be required here. So that's it for input. And then I have an output
10:43 also. So I could add an output. What happens at the end? And in this case,
10:48 it's going to get the script text, which again, the description full YouTube text
10:53 from that doc. So it's going to be able to with the file ID find a new document
10:58 and then extract the text and give the agent that text. That's what this
11:02 workflow is going to do. And if you set it to multi-line, it could write
11:05 multi-line. So, I'll set it to that. I'll press save. Okay. So, now I have an
11:10 input and output ready. Now, I'm going to press the plus sign and I'm going to
11:14 choose Google Docs, which again, there's thousands of apps you could connect to
11:18 this way. Google Docs is obviously one of them. And you have ton of different
11:22 actions within each document. In this case, I wanted to get content from a
11:26 document. It's going to see when a new document appears, and it'll get the
11:30 content from that. Then, it's going to ask you to connect. Obviously, it's my
11:33 first time using it on this account. So, you have to give a connection permission
11:38 to Google. So, I'll click on this and set that up. Very simple. Okay. I
11:42 checked on all the requests. And by the way, under the connections, you could
11:45 add new connections there anytime or remove any existing one. Always
11:49 available there. I'm going to do by mapping and I'm going to click on
11:54 document ID. Now, if you go to this tab right here at the end, you'll see your
11:58 custom system variables. That's your input that I selected myself. That's why
12:03 I set that up in the very beginning. It always appears here. So, I could just
12:07 select it here. Now, you could do it by dropdown, but I don't want to do it
12:10 right here by dropdown because I wanted to map here to what's happening in my
12:15 actual account. And the document ID came from that. I'm going to press save.
12:18 Okay. Now, I'm going to turn this into a scenario right now. So, I'm going to
12:22 press the plus sign here. And there is something called scenario. You could
12:25 look for it. It's this one right here. I'm going to select it. And under action
12:30 here, I want it to return an output. And if I click here, I want it to give me
12:35 the script text, right? And I have all these different options available on
12:40 what it can do. So the one I want over here is called text content. So I could
12:45 click on this right here. So this is going to get the script text. I'm going
12:48 to save that. Okay. I'll go ahead and save this here. Now, with this type of
12:52 scenario, it's also asking you to actually change it on demand because
12:55 we're pulling something from Google Drive here. So, I'm going to update that
12:59 schedule, and I'm going to go ahead and turn this on demand when I'm ready. But
13:03 right now, I'm going to go ahead and save it. So, we could actually do a
13:06 quick test. So, I'm just going to run it once to run a test. And to run a test,
13:11 it will for input ask me for the file ID. So, I'll just show you. If you go to
13:17 a Google Doc, you'll see your file ID here right after the D. So this right
13:23 here is the file ID before edit. So it's between d slash and edit. Again, that's
13:28 going to happen automatically. I'm just doing a quick test here to make sure
13:31 it's working. So I'm going to input that and press okay. And it's going to run.
13:35 And it's going to show you that it's working. So you'll see one action was
13:40 taken here. One action was taken here. And here under the log section on the
13:44 bottom, I could see exactly what happened. As long as these turn green.
13:49 Okay. And you get that each action was taken. This is working exactly like I
13:54 want. Now I could go ahead and turn it on demand. Now I'm going to give my
13:58 agent access to this. Right now this is just running as a scenario, but I'll
14:02 show you what happens once the agent has access to it. Okay, let's go back to the
14:06 AI agent tab here and we'll go back to our repurposing agent. So so far I just
14:11 created that new scenario. So I'm going to press add and I'm going to add it.
14:14 Now, when you add it, it will show up under the dropdown the way I made it.
14:18 So, I could click on it right over here, and I could give it a description.
14:22 Remember, with AI agents, always describe what this is supposed to do.
14:27 And this tool takes the file ID or document ID that it finds in a new
14:32 script scenario and uses it to pull the full script text. I'm going to go ahead
14:37 and add this one. Now, this agent has the AI model in the background, has the
14:42 system instructions, has my context with the tone, and has this one specific
14:46 tool. Now, this is already going to go to work, but I want to build the whole
14:49 workflow. So, we're going to add a couple more tools here. And this time,
14:53 we'll build some simpler ones. We'll just do the modules. With the modules,
14:57 it's a one-off action, so they're really easy to make. And we'll go ahead and
15:01 just build a module tool from this page. Very straightforward with these. The
15:05 first one I'm going to create here is going to be a Google Docs, but this time
15:09 it's going to be able to write a new one. Right, we're repurposing content.
15:12 So, I'm going to choose that one. And this is going to be able to create a new
15:16 document. So, a different action here. And my description is going to be create
15:21 a document for shorts and LinkedIn post text. It's going to be the same
15:24 document, but it's going to break it up. And I explained it in the system
15:28 instruction to kind of break it up into two different sections. It already knows
15:31 my connection because we already connected to our Google accounts. For
15:35 the name and the context, I'm going to ask the AI to decide for me here. And
15:39 then it's going to save it somewhere. So, I don't want to save in the same
15:42 folder because it's going to create kind of a loop. It's going to always see a
15:46 new document being created and try to repurpose that. So, I'm going to create
15:50 this in a new folder that I'm going to have which I already created called AI
15:54 created files. So, it will move into that folder instead of the folder my
15:58 scripts are living in. And that's it for creating new tools, these module tools,
16:02 because they're just one action. They don't need a scenario like the one I
16:06 just created here. They just literally do one thing, which is go create a new
16:11 document. Now, the next one, I'm actually going to have it post on
16:14 LinkedIn for me. So, this is really handy. We'll create a new one. This one
16:19 is going to be a LinkedIn and create a user text post. In this case, I want to
16:24 actually post it to my company page. So, create a company text post. And the name
16:29 of it is I'm going to put right here. LinkedIn creates an organizational text
16:33 post. And in this description, because I'm creating a doc that is multiple
16:38 different things for multiple platforms, YouTube short being one of them. This
16:42 says share only the LinkedIn specific repurposed content from the Google doc.
16:46 I'm going to go ahead and connect to my LinkedIn account here because this is a
16:49 new connection. I've only connected to OpenAI and Google so far. And I chose my
16:54 company page and I let AI decide the content here. And we'll add one more
16:58 here. Okay. Now, this is going to notify me on Slack when a new YouTube short is
17:01 ready, which again, this agent takes care of. It takes my long form, makes a
17:06 short form. Now, this is a lot more handy lately because we have other
17:09 researchers that are helping me with the script writing process. So, maybe that
17:14 way as soon as it's ready and they put it in my folder, this does the
17:18 repurposing. It does the posting on LinkedIn and it's going to let me know
17:22 when a YouTube short is ready. So, I'll create a new one. And this time we'll
17:26 choose Slack. You could send this to a group in Slack, by the way, too. Or just
17:30 this is going to ping me directly. Direct message. Create a send a message
17:34 on Slack. Again, I'll connect to Slack here, too. And it's going to be a Slack
17:38 user. And this is going to be a direct message sent to me on Slack on the Slack
17:42 channel I give it access to. And I'm going to let AI decide the text here.
17:46 I'm going to let everything be decided by AI. And I'll add. Okay, that's pretty
17:50 much it. You could build this a lot more. You could post to a lot more
17:55 platforms, repurpose this in a lot of different ways, but just to show you so
17:59 far, this is going to take a YouTube script that gets created in Google Doc
18:03 in a specific folder. Then it's going to actually repurpose it as a LinkedIn post
18:08 and a YouTube shorts script. Then it's going to post it on LinkedIn and it's
18:13 going to notify me on Slack that this is ready for me to now read a short version
18:18 of that script, too. Okay. So now you could go ahead and do some testing here
18:22 if you want on the right side, but we still need to create one more scenario
18:26 here. Okay. So let me go ahead and create it and explain what this is going
18:29 to do. Now this one is actually not a tool I'm making. I'm just making an
18:33 automation here for the agent to run continuously. So I'm just going to call
18:38 this right on top run agent. That's the job of this one. And again, if you build
18:43 more agents, name them accordingly because you may want to have multiple
18:46 different agents doing multiple different things. Now, this one actually
18:50 doesn't need input and output because we're not giving it to the agent. That
18:53 was more specific because we were giving something to the agent at the end. This
18:57 time, we're not doing that. So, I just need to start here with a Google Drive
19:02 connection. Let me show you why. In Google Drive, I wanted to actually find
19:07 out every time there is a new file. So, they have an option for that. You could
19:11 do it with Google Docs, too, but I like to use Google Drive for this. And only
19:14 because I wanted to specifically look inside of that folder. So this says
19:19 trigger when a file is created inside of a folder. So I'll choose that one. Now I
19:22 connected to Google Docs, but this is going to ask me to connect to Google one
19:26 more time for Google Drive access. Now this time I chose a very specific folder
19:30 here with the drop down. So I don't need AI to decide anything here. I'm going to
19:34 limit it to just one so you could always see the latest one. And I'll go ahead
19:37 and save that. And then you could tell it from now on or on a specific date.
19:41 Now, the next thing is I'm going to press the plus sign and I'm going to
19:45 choose make AI agent to then run that agent. Okay, this option right here. And
19:49 if I click here under the different agents that I have, I have one that I
19:53 created. So, I'll choose that agent here. This is how you set up a schedule
19:57 to run the agent. It could be by itself also. It doesn't have to be from a
20:00 trigger like this one. You could just have the AI agent running as your first
20:05 action here, your first trigger. So, you could see I have all the different
20:09 tools, scenarios, and modules. All those are there. I could always add more here.
20:13 You'll also see any MCP if you connected to external servers that way. Your
20:18 context is here. And down here, I'm going to go ahead and add a message. And
20:21 this is going to be my starter message as if I was talking to a regular
20:26 chatbot. This is the file ID for my new script in Google Drive. Repurpose the
20:31 content. That's what I'm telling the agent to do as a user. And here I'm
20:36 going to choose the file ID. So, that's going to add it here organically every
20:39 time it gets created here. And that's it. I'm going to go and save it from
20:43 there. Okay. Now that it's saved, I'm going to go ahead and turn this on. Now,
20:46 this doesn't actually have to run every 15 minutes, right? Because I don't get a
20:50 YouTube script every 15 minutes. I get it once a day. So, I could do this every
20:55 day at a specific time, right? Depending on what our schedule looks like. So,
20:59 this should be fine right now. I'm going to press save. So, as soon as a script
21:03 is ready, it'll know that. And usually by 3:00, we have one ready. So it'll run
21:08 exactly then. Now just for the sake of testing, I'm going to just change this
21:12 to run right now so I could show you the output that I'm getting. Okay. So it
21:16 went and checked for Google Drive. So one action ran there and it knows that I
21:20 created a new document. I actually just copied the script here and it created it
21:25 in the same folder as I have all my scripts. So now the agent is going to
21:29 work and after a few seconds it's done. So let's check its work. Let's go to
21:33 Google Docs first. Okay. So, here's the document it created for me. It created a
21:37 YouTube short. Now, with YouTube shorts, I actually like to break it down this
21:41 way into hook steps. That's where the custom instruction, that system prompt
21:45 that I gave it comes into play. So, that's why it's formatted this way. The
21:50 LinkedIn post is not formatted that way because this is going to go ahead and
21:53 post it. Now, I might want more line breaks, maybe a couple emojis thrown in
21:57 there, but that's where again my system prompt comes in. context is following
22:03 how my conversational style is and it's writing in that tone. And again, if you
22:07 build this out even further, you could have one for X, for example, for
22:10 Facebook. You could just build this out for other type of content you want to
22:15 post either manually or automatically. Now, on LinkedIn, here's my actual
22:19 profile, my business profile. One minute ago, I automated my YouTube script
22:23 repurposing with make AI agent in minutes. And this came exactly from the
22:27 script for this exact video. So I could show you what the output looks like
22:30 here. And I think it's pretty good. And if I go to my Slack over here, there it
22:35 is. My YouTube short script. I automated YouTube shorts with make AI agent.
22:38 That's the script that got repurposed. This link should take me directly to
22:42 this where I could go ahead and read that. And for the LinkedIn section too,
22:46 it showed me exactly what's been posted on LinkedIn. So, it's nice to get a
22:50 Slack message just to check what is posting. If I want to make some
22:54 alteration or fix my system prompt, the whole point is just run this and keep
22:58 improving on it till you have something you're really happy with and your
23:01 automation is running according to plan. Now, as you build more automations and
23:05 more of these AI agents, it gets easy to lose track of where everything connects.
23:10 So, make has introduced a new update called Make Grid that fixes a lot of
23:15 that for you. Now you could access grid on the left section here on the sidebar.
23:19 You could also look at it on these individual scenarios here. So I'll show
23:23 you this one right here. So you could see a nice visual presentation here and
23:27 you could move around your entire organization. So again, as this gets
23:31 more complicated, it's nice to zoom out to see big picture and zoom in. And you
23:35 could select by layer. So for example, I want to see how my data is being
23:39 transferred. It's going to show me that. If I want to see how my credits are
23:44 being used, I could see that by credits. Again, this is a brand new account, but
23:47 you could see how powerful this could become as you have more data, more
23:51 tools, more agent. You could get a really nice view to see how they talk to
23:55 each other here through all kinds of different ways. And I think if anything
23:58 breaks, this is a nice place to come back to to see how everything is
24:02 working. And this makes building AI agents a lot less of a mystery, right?
24:06 You have a lot more control here and you could see exactly what's going on. And
24:09 when you are done testing everything, this is again where you could run your
24:14 agent and turn it on and it will run based on the schedule that you set here.
24:17 Mine was daily. You could set it as often as you want here depending on your
24:22 use case. So hopefully that example of that make AI agent showed you a few
24:25 different ways you could create AI agents for yourself to automate all
24:28 kinds of different things in your business. Again, make sure to grab the
24:32 link in the description and take advantage of that free month of the pro
24:36 plan and start building your AI agent today. Thank you so much for watching
