you2idea@video:~$ watch ztyw2NFyS4E [48:31]
// transcript — 743 segments
0:01 Do you think you need to go to university? It >> could very well be that it's hard for
0:05 you to find a job. Everything that happened over the past one or two years,
0:09 there will be very few. >> Do you need to learn to code to succeed
0:12 in AI? >> I would say these days, everyone is trying to get into AI. But the question
0:18 is how do I need to learn how to code? What routes are there in? What ways can
0:21 I actually monetize it? And what does success actually look like? So, I wanted
0:24 to have this discussion with my good friend Dave Eblahar who runs Data
0:27 Luminina, a training program for people getting into Genai as a developer. and
0:31 also a founder of his own agency. So we have a lot in common there and this
0:33 discussion is really a debate for both camps of do you need to learn how to
0:37 code? Do you take the developer route or can you come through and learn no code
0:39 and build a successful business and create the life of your dreams that way.
0:43 So hope you enjoy this conversation with Dave. >> Dave mate what's up? What have you been
0:45 doing? >> What up Liam? Great to be here man. So primarily what we're focusing on right
0:51 now in data luminina is we're really building out the educational programs
0:55 that we have for all of the developers that we want to help in the AI space and
1:02 essentially help people all the way from never having written for example a
1:06 single line of of code all the way to the AI engineering that we do and then
1:10 also selling that as a service to clients for example as a freelancer and
1:14 then of course with the agency working on the client projects trying to scale
1:18 that up uh And that's constantly moving, right? Still trying to figure out how to
1:22 best go about that. The types of clients that we want to work with and the types
1:26 of like niching down we want to do with the agency. I think that's also
1:29 something you can probably relate to. Yeah. No, we can we can get into agency
1:32 niching a bit later, but I want I want to start off with what probably people
1:35 have clicked on here to uh to have a listen about, which is this, do you need
1:38 to learn to code to succeed in AI, which has been a pretty hot topic. I mean, I
1:42 got I got flamed a lot at the start of my kind of my career on on YouTube here
1:46 for I think even like before we were mates like you made I made the I started
1:50 talking about obviously the no code stuff and what we were seeing with like
1:52 client delivery. I think you made a video and you like said there's there's
1:56 some guys on the internet talking about this no code stuff but I think like
1:59 we've all kind of converged on the same conclusion that it doesn't really matter
2:03 how the [ __ ] you build the stuff like like with the agent builder coming out
2:06 and things like that it's very clear now uh what the what the final resolution
2:10 there was. But as someone who's literally training people into into G AI
2:14 development, you're probably the best person to talk to in this. But what's
2:17 your perspective on do you need to learn to code to succeed in AI? Um we probably
2:21 should get clear on what what success means first. So for for say your
2:25 students, what do you what do you class as as success? >> So when it comes to the students I work
2:31 with, success in AI primarily means look >> I want to learn how to code to probably
2:36 like land a job uh or start as a freelancer, do my own thing. But most of
2:40 them want to get a job in the field and that is as a programmer as an AI
2:43 engineer as a developer. So that's what most of my students would define as
2:47 success. But if we come back to look if you want to take part in AI, if you want
2:52 to ride this wave essentially, do you need to know how to code? I would say
2:56 definitely not. Even though like that's heavily that's my background, that's
2:59 what I specialize in. That's what I love. There are so many ways that you
3:04 can be successful within the space, right? I know plenty of people who make
3:09 a living like just going to businesses and helping just helping the the the
3:14 employees there get started with AI understand what what that looks like. So
3:18 for for the people like watching it really comes down to look at what kind
3:23 of level um do you want to work with AI? Do you just want to use the tools that
3:27 are already available? Do you want to learn how to use these models? Then
3:31 that's a level you can play the game at. or do are you like more interested in
3:35 this and maybe a little bit more technical and you feel like look I I
3:39 really also want to learn how to code and then that could be uh something for
3:43 you to focus on and that's also how you can be successful. So yeah long long
3:47 answer short you don't have to code in order to be successful in AI it's
3:50 something you can explore. I think one of the the good things that we're seeing
3:53 at the moment is the market really taking shape and over the past I'd say
3:58 12 12 months it's become very clear that there's a there's a development portion
4:00 there's of course like the no code development there's there's make.com
4:04 there's uh any of the I mean agent builder coming out now as well it's very
4:07 clear that there's a there's definitely a path into it that way and then there's
4:10 the more like I want to be an actual geni engineer and I mean these are the
4:14 people that we need at at companies and agencies like ours there's also the
4:17 consulting route that's popping up there's also the outright education like
4:20 you're talking about some people that you you already know, you can go into
4:24 companies and train them on how to use JGBT. I I still think that's a massive
4:28 opportunity. I've had people who are who are good mates of mine getting into
4:31 doing workshops. I think if you're looking to come in like the original
4:35 route into AI was was through the the low code or or to actual development and
4:39 now there's like I think you kind of flip it on its head and go in through
4:42 the workshop route, go start with education and then move down to offering
4:46 consulting after that and then eventually get into development at the
4:49 end of it. So you used to be able to only go through the development portion,
4:52 but I think now you could come down from the top and start off as nontechnical
4:55 and if you're talking about like success, if it's monetary success,
4:59 there's so many different ways to do it now. So there's definitely still a
5:02 massive need for the for the AI engineers and that's only going to
5:05 increase. Like we're we're always hiring as I'm sure sure you are for good
5:10 talent, but the the definition of of success is is really broadened a lot.
5:13 Obviously you teach the the genai development actual actual codebased. Um,
5:17 do you think it's necessary for those kinds of people to also pick up the the
5:21 lower code stuff and be kind of cuz we have this as well when we're hiring for
5:25 our developers at Morningside. We will also say, hey, it's like it's a it's a
5:28 nice to have if you're also familiar with that whole no code stack because
5:32 it's just like in a lot of cases it's much more expedient to to build on that
5:36 or we might have a client project that is already using that or we want to
5:39 build something ourselves. So there's like two sides to it and obviously the
5:42 the ideal developer would have both. So what do you think? Okay guys, very
5:44 quickly two things from me. If you're a business owner who's interested in what
5:47 generative AI can do for your business, you can get in touch with me and my team
5:50 at Mornings.AI in one of the links in the description below and we can start
5:53 your entire AI transformation process going all the way from the education and
5:56 training of your staff to the identification of the best AI use cases
6:00 for your company all the way through to development and beyond. We have worked
6:02 with some of the world's biggest sports teams and also publicly traded
6:05 companies. So rest assured, you are in good hands. If you're an aspiring
6:08 entrepreneur and want to start your own AI business and you haven't already
6:11 joined my free school community, it's down there in one of the links in the
6:14 description below. has my full free course on how to start your own AI
6:17 agency as a complete beginner. And you're surrounded by over a quarter
6:20 million people who are also striving towards the same thing. There's no
6:23 better place on the planet right now to be surrounded by like-minded people. And
6:26 you get free weekly Q&A with me where you can ask questions directly to me
6:29 about how to start and scale your business. I'll see you in there. I
6:33 totally agree with you there. We in our agency had to learn the hard way. So,
6:38 um, a lot of engineers of course super technical, right? They want to solve
6:42 most of their problems through like the the technical knowledge and the tools
6:46 that they have, right? So, similar to when you have a hammer, everything looks
6:50 like a nail, right? But then stepping in and like zooming out a little bit and be
6:54 like, look, this is actually a problem. Like you don't need to custom build for
6:58 this. Like you don't need to spend like two, three weeks, maybe even a month to
7:02 set this up and then have the whole like production pipeline. This can literally
7:06 just be an NAND automation and we can set it up in a day. Even if you are
7:10 super technical, even if you're an engineer and that is what you do,
7:14 knowing also how to use these low code no code tools and knowing how they work
7:20 and also being able to u uh clearly educate a client and say look you're
7:25 asking for this but it would actually be better for you if we do it that like
7:29 this long term. This is also something I'm uh teaching in my programs right now
7:34 just so um that also the clients that also work with our students have a
7:38 better overall experience so you don't end up wasting like tens of thousands of
7:42 dollars maybe on a proof of concept that in the end doesn't work that maybe could
7:46 have been validated with a quick NAN build. >> Yeah 100%. Um and with with your
7:50 students that you're working with why don't you if you could just give a rough
7:53 rundown if someone else was wanting to take that path and get into the point of
7:57 like basically AI freelancing. um why don't you give your your route through
8:00 that you're pushing them pushing them through and then I can kind of give uh
8:03 and and some examples as well of of sort of the other route if you're completely
8:06 nontechnical now and you're wanting to learn AI I can I can kind of give what
8:09 I've seen works on on mine. >> Yeah, sure. So I can quickly walk you
8:13 through kind of like my path. Um and then what I do right now is I I teach
8:17 the exact same path that I went through but then just in a way I could just
8:21 >> not in 10 years or whatever. >> Just not in 10 years. So I started I
8:26 started studying AI actually in 2013 already. So this is literally over 10
8:29 years ago. So did the bachelor, did the master uh yeah technical training and
8:34 then out of uni I started um I started working as a data scientist and I
8:39 already immediately got into freelancing just just came came out of nowhere on my
8:42 path like in the beginning like the first one to be honest was quite lucky
8:46 and then from there you just like built built the skill built the muscle and you
8:50 essentially know how to find the next project and the next project and that's
8:54 essentially how you roll into it. That's how I got there. First it was in data
8:58 science and then when chat GPT was released I immediately saw look I need
9:03 to go all in on AI this on genai I should say I need to go all in on geni.
9:06 So that's what I did with it with the business. And at that time um before
9:13 this chat GPT boom and when I was just starting out with YouTube, what I
9:18 primarily was doing, I was mostly going from long-term contract to long-term
9:23 contract. Because one of the things that you should understand about freelancing
9:26 in the tech industry. So this is if you have technical skills, if you are a
9:30 programmer, it's not so much that you jump from like a 3K project to a 5K
9:35 project and you're juggling three to five clients per month. Like you can do
9:39 this, but as a freelance in the tech industry, like 80% of them, they just
9:44 have one or two clients at the same time and just cruise three months, six
9:48 months. I even had contracts that were just a year long and they're not
9:52 full-time. And then on top of that, I would stack other uh projects as well.
9:56 And there are actually people here who are paid to find people like you and
10:00 place them inside companies where you can just cruise for 24 hours per week or
10:04 32 hours per week. And then that is a way into freelancing for you. So that's
10:08 kind of like a stepping stone that I did in the beginning. And then um as uh Chet
10:13 GBT was getting bigger and bigger at that time, I was already I was I already
10:17 had my YouTube channel. Creating content is of course another great way to like
10:22 get yourself on the radar. I don't have to tell you this. And with that, you
10:26 also find that over time, if you do it right, you'll get inbound leads. So,
10:30 this is now where you have these longer term contracts which are paid really
10:34 well. Like if you're an engineer, like hourly rates, €100 or 100 USD, 120, 150,
10:43 and if you have like 24 22 or 20 32 hours per week, you can make great
10:47 money. So that's all great. Then when you stack a personal brand on top of
10:50 that and you also start to get inbound leads and you with those you focus more
10:55 on the on the quick projects that you maybe do via a uh for a fixed price,
10:59 right? That is essentially the the secret sauce in the mix that we teach
11:03 inside our program. So that hybrid approach where on the one hand you have
11:07 the security as an engineer or developer and you stack smaller more higher paid
11:10 projects on top of that and that's how we find that students have really high
11:14 success rate. It's really fun. the work is dynamic. Uh, but you never have have
11:20 to have that feast or famine life cycles that you can have, right, as an agency
11:23 owner or as a freelancer. >> Are you saying that everyone should do
11:25 the the content thing? Cuz that's something that I've I've struggled with
11:29 with um helping my students. >> You should try content and see if it's
11:32 for you. See if you like it. But it's not for everyone. It's just not for
11:38 everyone. But you you don't know that until you give it a try. So this is and
11:41 this is then also like kind of like tricky to do, right? is you need to get
11:45 in a couple of reps because in the beginning everyone's going to suck,
11:48 right? In the beginning, no one is good at this and you need to get better at
11:51 this. You also need to have just some kind of like it factor, right? Some
11:55 people just have that and then you see the channels pop up. That is just
11:59 something where you just got to you just got to be really honest with yourself
12:03 and you got to put in the work. You got to try and like make a couple of videos.
12:07 See if you like it. See if you get some traction and also like critically look
12:11 at at the videos and actually like look it is this adding really any value? Am I
12:15 really helping people? What are what do the comments look like? And if you get
12:19 some early traction there and you really like it, that could be a good signal
12:24 that you should continue with that. If you like absolutely dread it and you
12:29 tried like three, five videos and you're like, I hate this. I don't see any
12:33 results. It might be that content is not for you. And it might be more so that
12:38 for example as an engineer, as a technical person, you are way better off
12:43 maybe for example going on a platform like Upwork and being more behind the
12:47 scenes and just bang out proposals there and make sure you create a killer
12:51 profile there and build up the reputation there so you never have to
12:54 worry about any sales or marketing and clients just come to you. So that's
12:57 another way that you can create inbound leads. there are more uh alternatives
13:01 that you can get into >> the pathway into it. We've got obviously
13:04 there's there's like a Python base and stuff and I assume there's like do you
13:09 put your guys through a mandatory set of like fixed projects like hey you're
13:12 going to run into a rag system you're going to run into like a a chat 2
13:16 database system like what's is there obviously a Python base probably if
13:19 they're learning that initially and then going through to uh knocking off a few
13:23 of those initial builds like what would you say if you had to have like the main
13:28 three to five uh use cases that if you were to learn to code uh you would need
13:32 to to know as a GI developer. Great question. So when um when it comes to
13:37 essentially building Genai applications that can actually make it to production,
13:42 I would say there are six key components that you really need to understand
13:45 beyond just the basic Python skills, right? So first is really about the uh
13:51 really the the fundamentals that you need to understand about the the
13:53 language models that you're working with. So this is everything that has to
13:58 do with like tokenization, prompt engineering, how do these models work.
14:02 This is the entry point. So you start to get a fundamental understanding of look
14:06 we have this medium that is a language model and we as engineers we can program
14:10 around this. This is also where you get into all of the SDKs for all of the big
14:15 models right so open AI entropic understand how how they work understand
14:18 what function calling is understand what structured output is uh how to use tools
14:23 in there so really these fundamentals then the second thing that you really
14:29 need to understand is um system design. So this is now knowing that you have
14:33 this this medium that is a a large language model and how to look at the
14:38 business process and I always recommend like really creating diagrams and going
14:42 to the whiteboard >> and then designing the architecture and
14:45 designing the system. This is also sometimes referred to as a cognitive
14:49 architecture. So can you draw out on the whiteboard what the business process is
14:53 where you just need like simple programming logic and where you maybe
14:58 need large language models because all really AI engineering and GI engineering
15:03 really is about is about essentially taking a problem and finding a small
15:07 enough component where you want to pluck in the language model. You don't want to
15:10 solve big problems with large language models because that can get really
15:14 messy. The third thing that you need to uh essentially that we also walk people
15:20 through is getting to the core of okay look I can uh take this model I can
15:24 build stuff with it but how do I piece it together in something that we would
15:28 call a production ready application right so this is your endpoints your
15:32 databases how do you stack that together using ideally using docker that's what
15:36 we recommend and you package it in such a way where now it becomes an
15:40 application then you're already uh really far ahead in terms of what you
15:42 can And then if you want to get even further, what you also need is you need
15:49 a solid solid understanding of rack and all the like concepts around that. So
15:53 retrieval augmented generation. So factor databases, factor pipelines, how
15:58 you can take uh document processing pipelines and then also advanced
16:02 algorithms that you can use to uh improve your your retrieval ranking.
16:07 >> Yeah, reranking etc. So all of these concepts around rack I would say that's
16:11 also a core skill where if you want to get into any job uh whether it's
16:15 freelancing or full-time like you need to know rag because otherwise the
16:18 interview will be kind of awkward. Essentially the next step after that is
16:23 this is probably the most important one everything that has to do with
16:28 monitoring evaluations and guardrails because these applications are messy.
16:33 The thing the one of the challenges with Gen AI and with AI engineering is as
16:37 soon as you launch something whether it's a proof of concept and MVP it's
16:42 never going to be 100%. It's going to be 60 70 80% and this is tricky because for
16:50 some use cases 80% might not be enough and it might take you for example like 1
16:55 2 3 weeks to get to 80% and it might take you 6 months to get to 85% and it
17:00 still might not be enough. So you need to have a really good set of tools and
17:05 skills to understand monitoring. So we for example use lang fuse uh for that
17:09 internally. You need to understand about guard rails, how to set up evaluations.
17:14 And then step number six that you also need is uh a solid understanding of
17:17 deployment. So your deployment strategies, you take all of all of the
17:21 things that you did before that and now you can actually put that onto a server
17:25 or deploy it somewhere and now it can actually start to generate value. So
17:29 that's the road map. we uh we take people through um and also all of the
17:33 things that I had to learn over the over the last years which we now use to help
17:37 yeah build and deploy all of the projects for our clients. Well, I'd be
17:41 interested to to know your thoughts on I mean, you talked there about the the
17:44 difficulty and the success rates that that you encounter in pretty much any
17:48 any tune AI project and and we know this very well at Morningside as well, how
17:51 much that last like 5 10% can be sometimes and on some use cases it's
17:55 it's okay and like you can have a little bit of particular if there's like a
17:58 human in the middle or or a staff member using it, but as soon as that's going
18:02 out to the prospects um and the outer world, then then you're in in trouble.
18:07 But with this new agent builder and what OpenAI has released there, obviously
18:11 multi- aent and I mean that just makes me like sort of shiver as a as an agency
18:15 owner because I know like a single agent is often difficult enough to get working
18:19 let alone multiple agents talking to each other. So and baked into that we've
18:23 got observability, we've got evals and things like that. So what are your
18:26 thoughts on on this new platform? We're already kind of looking at it for some
18:29 of our builds that we've got coming up. Um and just try and say, hey, how far
18:33 can we push this thing? um can we replace them with our custom code with
18:36 it? Uh what are your thoughts overall on the on the agent builder and how it fits
18:39 into into the agency landscape and just to develop a toolkit that way?
18:42 >> Well, it it's one of those things it looks very promising. I haven't really
18:48 like fully uh like build out an entire project with it, but based on what I've
18:53 heard is like so it's early, right? And OpenAI has a tendency to release stuff
18:58 that in the demo looks really good like like the assistance API was also one of
19:01 those things where we thought oh yeah we don't we can just use the assistance API
19:05 right we it can take care of rack it can keep track of our our conversations in
19:10 like the real world like no engineer is is using that so I feel like the the the
19:15 agent the agent builder is in a similar stage right now where it looks promising
19:21 but it's still the thing is you are you're building an abstraction a
19:26 a visual interface over the core logic where at some point you're going to hit
19:32 a limit. So I think this is still going to be super valuable and for like but
19:38 similar in a way how NAN is super valuable and make.com is super valuable
19:41 and all of these tools are super valuable in their own way but at some
19:46 point you're you're likely going to run into a limitation where yeah now we need
19:51 to tweak the performance even more and it can get tricky. What I like about
19:56 agent kit, agent builder in uh uh specifically is that it's like fully
20:02 like LLM geni first, right? If you look at NAN, make.com, Zapier, they are they
20:07 were all like low code no code like stacked AI on top of it. So what you see
20:13 in the design of the system is you see all these common patterns that we need
20:17 when we're trying to build agent applications. So like guard rails like
20:22 completely built into it because it's really important and making that really
20:26 easy and of course like changing all the different models and prompts and all of
20:32 that. I like how LLM and Genai native it is. Um and I think it's a really good
20:37 tool but I think ultimately you're going to run into the same issues that you
20:41 have when you're for example using N8N with more complex builds that at some
20:45 point you're just going to hit a wall. >> Yeah. I mean when I was looking at it,
20:48 it obviously looks looks nice and they have made these kind of big big promises
20:52 before. Um I I think it's interesting from a from a perspective I see them as
20:56 kind of being the the apple of AI automation at this point. You know,
20:58 they're trying to make the closed ecosystem where like you know you're not
21:01 going to be plugging in all these different models. You lose the ability
21:05 to be model agnostic, but what you get in Exchange is just like ease of use,
21:09 ease of experience, like a nice customer experience and user experience when
21:12 you're setting it up. Um, and then all these little like quality of life things
21:15 that are sort of native to it because it's GI first. So, I'm excited. I'm
21:18 going to be jumping in more and seeing what I can I can cook up and maybe
21:21 putting a video out on it. But, I was kind of shocked when it came out to be
21:24 honest. I didn't I didn't see them going that way. But, it's it's just another
21:27 example of like when the assistance API came out, they nuked a whole bunch of
21:31 startups there and they've done it again. Right now, these are all early
21:36 stages, right? But what you'll find often is that like Chad GBT remains the
21:41 center of it, right? it usually remains like the nucleus which where it can
21:45 really easily integrate with with chat GPT who's like everyone is using right
21:50 so I think having that power and then on top of that like making all of these
21:54 other tools better and better and better where that that could be really where
21:59 you get that Apple effect that lock in where you feel like yeah okay I could
22:03 use that other tool it might be like a couple% better but it's just so easy
22:07 that I have everything in here why would I even bother >> that's the allure right I think It's I I
22:11 call it like Apple versus Android where are you going to maybe take a little bit
22:14 longer to set up and I think once you layer on top something like these these
22:18 like vibe automation things as I'm I'm calling calling it now where you can
22:20 like if you go to NAT and they've got for some people you have access to the
22:23 to the like text interface where you can go text to workflow. I think like the
22:29 ease of setting stuff up is going to be insane when you have a like an assistant
22:33 in there on the agent builder where you can like type into it, hey, I'm building
22:36 this and instead of actually having to move things around manually yourself. um
22:40 you're able to to prompt in to get it to build it for you. So, uh that that does
22:44 open a big question which a lot of the agency owners on here would be would be
22:47 eager to know like this seems like development development is kind of
22:51 racing to zero and it's like a business is going to start to do this and my my
22:54 push back on that is like no this is a dev day. It wasn't a business day like
22:58 it's still developer tools like business are not going to be hopping on there and
23:01 building their own multi- aent systems. So, it's still uh very much very much in
23:06 our in our uh court. But with the development becoming a lot easier,
23:09 there's questions around like is this going to lead to a massive influx of new
23:12 people in there? And again, I still think like these no code platforms been
23:16 around and they're probably a lot easier for most of the basic use cases. People
23:20 have had plenty of time to get into it. Um, but there's sort of the broader
23:23 question of of what happens to development long term as these things
23:26 get easier. So, what are your thoughts on on Genai development with the rise of
23:30 these no-go platforms like viable automation becoming easier and easier
23:33 and how that's going to affect the uh the agency market? So first let let's
23:39 start what we already see happening in in the big corporates and the
23:42 enterprises right and then we can get down to kind of like the agency uh level
23:48 where what we're already seeing is that kind kind of like the paradigm of really
23:53 big organizations with huge engineering teams. We already see that shrinking
23:57 because they're just trying they're just figuring out look if we have like a
24:01 couple hundred instead of a couple thousand like really good engineers with
24:07 AI assisted coding tools we can be much more efficient. So there definitely
24:12 things are changing and yeah we're not really sure where this is. Like on the
24:16 one hand like every other quarter you get like these crazy new up news updates
24:20 where it's like oh this company like fired like half of their engineers and
24:23 then another quarter is like oh they hired them back. So like we're still
24:27 like trying to figure out but I I there is definitely a trend going on like
24:32 developer productivity has increased tremendously. I think out of I think out
24:39 of everything where AI can add value and add productivity right now, it's
24:43 probably the developers that benefit the most from it in terms of like pure like
24:47 hours saved on a monthly >> ROI. Yeah. >> Clear ROI because it's just so good
24:54 there because LLM's and and and AI is like coding is such a nice and defined
24:58 problem, right? You have a problem, you create a code, you run it. Does it work?
25:01 No. Okay, you iterate on the error and >> it's all basically text base, right?
25:05 like a >> it's such a nice environment uh uh to uh to use these language models in and it
25:13 is now really up to up to us up to the world to figure out okay what's going to
25:17 be the cursor what's going to be the clawed codes but then for other domains
25:21 right for the medical domain for for law whatever and that that's a trend that
25:25 that we're seeing there but then if we boil it down to like the the smaller
25:29 organizations what I think is going to happen is we will see distribution of
25:34 all that development talent. Well, where before it was more concentrated
25:40 um in larger organizations where huge development teams would work together
25:44 and a lot of companies don't need used to not need a developer.
25:49 Many industries like you don't need a developer. Right now is like every I
25:55 feel like every company if you want to stay competitive you should have AI on
25:58 your radar and you should do stuff with that. Even if you're just like a con
26:03 construction company, even for like the admin administrative process or
26:06 something like that, AI can probably help you there. So, what you'll probably
26:11 see is that you'll have more and more smaller companies that are not going to
26:15 hire a full-time developer because that doesn't make sense, but they do want in
26:19 on AI. So, then they're going to look around them and that's where the
26:23 agencies and the freelancers can come in. So, I that's that's that's a trend
26:29 that I see. So when we are in this like agency freelance tech bubble right where
26:32 we see that there are so many people like jumping onto this but if you look
26:36 outside our bubble there are also so many businesses who are starting to jump
26:41 on on AI in the nearshort term thinking like next 1 2 3 4 5 years and who knows
26:47 where AI is then I I still think there is plenty of space for developers at any
26:54 level whether you use n10.com make or you become a full-blown AI engineer
26:59 there will be opportunities for you, but you need to be more strategic about it
27:04 because it could very well be that if you now go to uni, you do computer
27:08 science, you specialize in AI engineering, it could very well be that
27:12 it's hard for you to find a job if you go the traditional route and try to uh
27:17 uh work for the top companies because they're just like rep prioritizing as to
27:21 what it means to have uh a big engineering teams. Do you think you need
27:26 to go to university at this point to get into to to succeed in some way? What
27:29 would you recommend? They do the kind of self-taught scrappy route or they go to
27:33 university, do the computer science and they add AI on top of that. What do you
27:37 think is is is the best best route at this point? >> With the knowledge that I have right now
27:41 and if I look what I learned in uni, there are so there are so much faster
27:46 ways to to get up up to speed into what's going on. And on top of that, if
27:50 you go to uni right now, all of the stuff that you'll learn on AI, even even
27:54 you like computer science, artificial intelligence, that like curriculum will
27:58 be outdated because they they cannot keep up. Like everything that happened
28:02 over the past one or two years, there will be very few like uni studies that
28:08 uh teach what what really can get you a job right now. But that said, having
28:12 like the credibility, having the university paper, it still is a form of
28:18 credibility, right? And if you want to get into certain fields or you want to
28:23 get into certain companies, it might not it might still be required. So, you
28:28 won't even like really get on the radar if you don't have a degree. That's still
28:33 the world that we that we live in. But is it a is it a good way to learn AI?
28:38 Definitely not. just go to YouTube and like watch what the the creators are
28:42 doing there. Watch the the the latest news and how they are building with
28:45 these models. So that's how you learn fast and then coming back to do you need
28:50 it. So it it depends on the route that you want to take and if you're like look
28:54 I'm also totally okay with maybe like working for like smaller companies
28:57 starting my own thing then they don't care about your degree and your
29:00 credentials. They only care if you care about what you >> done. Yeah.
29:05 >> Yeah. Yeah. The experience talks 100%. And um it just goes to show if you're
29:08 looking at like the people like with programs and and selling courses around
29:11 this stuff like even in the online space it's hard to keep up and like even for
29:15 you you can be you're very very very like on top of things but as soon as you
29:18 make course material 3 6 months later you got to make new stuff like it even
29:22 for people who are are literally moving at the at the fastest pace with the
29:26 stuff. It's still hard to to keep pace. So you can't expect some slow clunky
29:29 university to be able to do it. The way I see it at least as someone who like I
29:33 dropped out of uni and I'm glad I did. What do you think is going to get you
29:35 further in your career? Is it like four years, 5 years and a piece of paper that
29:39 gets you that credibility or four to five years in the trenches building it
29:42 and working as a freelancer or building your own agency and stuff like that? So,
29:45 and I think if you do it right, obviously the four to five years spent
29:48 in a business like we're I'm only we're only getting into really like year three
29:52 here and look at look how far both of us have come. So, uh I I know where I I
29:56 stand on that for sure. >> I share this perspective, right? So, you
29:59 can follow your old traditional model and it it it might still work out,
30:03 right? that is still a proven path that you can take, but you don't have to in
30:08 today's day and age. And if you want to keep up, uni is not the place to go to.
30:12 And there are plenty of opportunities if you can just if you can build stuff, if
30:17 you can help people, which now is easier and easier because here here's also the
30:22 thing. When I was in uni and when I just started working, AI and data science was
30:27 only really for the big enterprises who had like big budgets, big data sets. We
30:31 would train custom machine learning models. It would take like months to get
30:34 to a model that we could put into production all like very long big
30:39 enterprisey kind of like projects. But right now AI like I've said literally
30:45 every b business can benefit from that. And knowing that opens up the door for
30:50 so many opportunities if you just learn how to like work with these tools and
30:55 actually create useful automation. So don't just like study the material, go
31:00 and build stuff like and first try to try to solve your own problems, right?
31:03 That's how how I always like to learn. Like for example, when there's something
31:07 new, let's try and automate something in my business using some of these new
31:11 stacks and then you you it starts to click and then once you prove that there
31:15 there's probably someone else on this planet that is willing to pay you for
31:19 that uh to help them solve a similar problem. >> You've run through the the the coding
31:24 route to go and that that you recommend. Uh, as for if you're if you're
31:26 non-technical and you're watching this and you're wondering about like, oh,
31:29 what's what's my best way? And if I don't necessarily think I've got the
31:32 I've got the got the skills or or I don't really see myself going into the
31:37 full code way. I have seen like dozens if not hundreds of people now, they go
31:40 into the the low code route, no code route, and they just kind of like bash
31:43 their head against it until they figure it out. And it's completely possible.
31:46 Like we have uh I had Henrik on the channel recently and he he joined my
31:51 program uh like a year and a year and a half ago, maybe maybe even more now. Um,
31:55 and for the first like six to 6 to 12 months, he came in completely green,
31:58 like absolutely no business or or AI experience. And now after bashing head
32:02 on.com for I think it was like 6 months before he really started to get the
32:05 knack of it and started landing clients. But 6 months after that, he started
32:08 working with voice agents. I connected him with Giannis and now they built a
32:11 really successful agency and a program as well. And he's one of the one of the
32:14 leading voices in terms of practitioners actually out there delivering voice AI
32:18 for businesses. He's there delivering like multi multi-f figure projects for I
32:23 think they had like a billion dollar uh client. So it's completely possible for
32:27 you to go that route. And it's just a case of I mean I had Mckll on here
32:30 recently as well. I'll link that uh in the description, but he was completely
32:33 non techchnical as well. He's like I'm just going to learn make.com as my base.
32:36 And I think as a meta skill learning automation via nad or make.com probably
32:41 push beginners to to make.com just to start. It's a little bit more beginner
32:44 friendly. But you can just sort of bash your way through it and finding okay
32:47 personal problems. There's so much information on YouTube now for it.
32:50 There's I've got a full course on on learning make automation and
32:53 particularly the AI portion of it. So I'll also link that down below. But all
32:57 the information is out there. I think people just need to be a bit more
33:00 patient with themselves. And like you're learning an entirely new kind of
33:03 automation development skills here from the ground up. It's not going to happen
33:07 overnight, but it's definitely possible within 3 to 6 months of just working
33:10 away. If you do some personal projects, do it for your friends and family. start
33:13 to reach out to people around you. Hey, I've learned this AI automation thing.
33:16 Uh would there be anything that you're interested in maybe having a chat and
33:19 see if I can help your business? I'll do it for free. And then from there, you
33:22 can go to communities like mine and yours and start posting. I've had so
33:25 many people now who are in there and they're making posts and saying, "Hey,
33:28 look, I'm available for work. I've done sort of these these few things. I can
33:30 have a chat with you and let you know if uh there's anything I can help you with
33:33 offering work for free." And then from there, they start getting paid clients
33:36 off the back of that. I see even like another layer coming out at the moment
33:39 in terms of like succeeding in AI where there's these like service delivery
33:42 platforms that have abstracted it one layer even further where you've got I
33:45 see you having custom code as the bottom layer and you kind of abstract that into
33:49 no code and then now instead of like having to learn like six different no
33:52 code tools and make like have a kind of flexible skill set you can abstract that
33:56 even further into these like targeted platforms kind of like what Simon has
33:59 with with Booked in that takes a lot of the complexity away of connecting six
34:03 tools to just being able to say okay I'm going to build a inbound voice agent or
34:06 receptionist on this platform and it takes away a lot of the complexity for
34:10 me. So that's a really really good time for people to be getting in and getting
34:13 onto onto those platforms and that's just the development portion. If you
34:16 want to go the AI consultant route, that's a thing. If you want to go AI
34:18 workshop route, that's another I'm so blessed to have been able to get into
34:21 this at the time we did and I still think it's ridiculously early. Uh I'm
34:25 I'm almost close to saying like >> it's pretty hard to not like get some
34:29 success in this in the space if you just apply yourself, you keep your ears open,
34:33 you are participating communities. literally I haven't seen a safer bet or
34:37 a sher bet or like a guaranteed bet in terms of entrepreneurship um that I wish
34:41 I had when I started. Man, I wish I started on this stuff. >> I was like bashing my head away on on
34:46 like e-commerce stuff and just eating [ __ ] for so long and losing all my
34:49 money. Um this is such an incredible time for particularly young people to to
34:52 start the careers. >> There's one more thing right now which
34:57 is also was also uh worth something uh looking into right now which is I'm
35:02 curious to hear your take on it. So you know right now MCPs are like everywhere
35:06 right? So MCP servers where you have all kinds of tools that are available right
35:10 now and more and more companies are creating MCP servers for their tool
35:16 right and you can hook them up to either chatpt or to cloud desktop and one of
35:23 the things we are exploring right now is instead of like doing builds kind of
35:28 like from top down. So you come in, you solve a big problem and then you embed
35:32 that into the business. You kind of like work your you work your way up and you
35:38 start where people are already using AI because everyone in almost all company
35:42 like you have some teams already most of the people are already using chat GPT
35:45 they're already familiar uh with those tools. So then instead of coming in and
35:50 doing a build for example either like custom or through make.com or nadn to
35:55 automate a portion you make it available as a tool within an AI chat application
36:00 that they're already using. And this way you also automatically embed the human
36:05 in the loop. And now all of a sudden you can let someone like let's say a sales
36:10 rep you can say hey what are my um what are my appointments for today? Use the
36:14 MCP server from the calendar. It loads in the context of that day. Okay. Now I
36:18 see okay I have a I have a call with John at 10. So now all in chat GBT hey
36:23 let let's actually pull the latest information from the CRM like for John.
36:28 And now you actually get like a report for the sales rep all in chat GPT. And
36:33 this is just different way of trying to creating these automation doing it in
36:38 inside these applications which I think even if you're a beginner and you're not
36:42 that technical hooking up those MCP servers through tools like chat GBT and
36:46 cloud desktop anyone can learn that. >> Think of that as like a service like
36:49 I'll go into your company and I'll set everyone up with the right tools. I'll
36:53 I'll hook up the MCPS to their uh to their claude and then I'm going to show
36:55 them how to do it. Is that is that the service? >> For sure.
37:00 >> Yeah, for sure. I really like that genius >> for you. Like this is a sick surface
37:03 because you can come in and you can can create all kinds of offers around this,
37:07 right? I could do 10 tools, tendium CPU servers, this that you can you can you
37:11 can really package that and >> what essentially the pitch is here is
37:14 say >> let's say look people inside your company they're all already using CGP
37:20 right yes they're already using CHP okay that already makes them a certain
37:24 percentage more productive you like you can find the latest information on I
37:27 don't know the exact numbers but essentially like the pitch that you have
37:31 is look in two weeks I can make all of your employees who use CHBT I can make
37:36 them 5% more productive by integrating custom MCP servers specifically for your
37:39 business. And >> that's on the low end. That's like absolutely on on the low end of things.
37:44 Um I mean that's that's just genius cuz you're just like the interface is always
37:48 a tricky part, you know, like you can build to the back end and stuff and then
37:51 >> worry about that. You can rely on R&D budget from OpenAI to handle this for
37:55 you. >> Yeah. No, that's genius. That's genius. Um I I think just generally I'm I'm
38:01 really still waiting for someone to do it. We have our own education uh like
38:05 workshops at Morningside that we do. Uh but in terms of building out an entire
38:09 like I still think if you did a a chatb course for businesses and you broke it
38:12 sort of general chatbt training and then you broke it down by like uh by
38:15 department. So like okay if you're in the marketing department here's a few
38:18 different like here's the specific skills of how you can apply deep
38:21 research or how you can apply uh the image generation or you can apply sore
38:24 and things like this and you break it down within each department. I think if
38:28 you threw on top of that like I'll set up I'll set set up the MCPS for each uh
38:33 each person's uh chat or cloud instance I think that would that would absolutely
38:37 cook. So I really think training that that seems to be the lowest hanging
38:40 fruit in my eyes is just teaching people how to use these tools is already the
38:44 data on how much more productive uh people are when they're using it. It
38:47 should be should be an easy sell and also it's probably a lot more lot more
38:50 scalable than like outright dev. Yeah, the MCP thing smart, bro.
38:54 >> Yeah, it it's a good one. It's like like it like you don't need to worry about
38:58 the UI that's covered for you and these MCP servers are getting get better and
39:03 better and as the models get better. This is also one of one of the things as
39:07 the AI models get better your solutions also get better. So it's just is like
39:10 one of the things you just bet on OpenAI and Enropic to like go crush it and then
39:14 you can build your service around that. We see a lot of opportunity in there and
39:17 we're just kind of like right now waiting for like the right client, the
39:20 right use case where they maybe want to custom build and we say look I think we
39:24 can do this do this just through MCP servers. >> I mean there's the there's a next step
39:27 after that which is like trying to build the white labelled version of that and
39:31 doing it through through the API. Um which I still think is a massive
39:35 opportunity. I I've seen some some startups appear that is kind of like
39:38 they call like an AI platform for your company and I think this is like a a
39:42 great idea for someone to to build out. probably a bit on the like sort of heavy
39:46 side, but when I think about what AI agencies can be upselling to their
39:51 clients if you have this like hey all of you you already got like 90% shadow AI
39:54 usage they're all using it they're all sending your data through the through
39:57 the chat to you without knowing there's a big like risk of leakage there and and
40:01 and sort of compliance and stuff like that there's that risk how about we set
40:05 you up a custom platform these things exist already and it's you can have
40:08 access to all the models you might be able to like query them side by side and
40:11 then you can start to add on like okay the co-pilots and the agents that you
40:14 build can start to be put in there as well. And basically you can add like
40:17 permissions and you can start to analyze how people are using I think just
40:20 analyzing how say you had that MCP thing set up as well knowing how your your
40:26 your team are using using chatbt and the things they're using it for that will
40:29 sort of identify the use cases for the agency to build out. So if you had a
40:32 platform, it's like, okay, everyone are going to use JPT and Anthropic and
40:35 Claude and things like this all through this platform and you're able to run
40:38 analysis off the back of that for the business owner saying, "Hey, look, looks
40:42 like Sand Sandra and and Jimmy and the marketing department are using it for
40:45 these specific tasks. We could potentially like mine like the inside
40:48 out of that and build something." Um, that's kind of that like bottom up
40:51 allowing them to figure out the the path of least resistance.
40:54 >> Yeah, it's great. You can also like see who the power users are, right? So you
40:57 can see like essentially what they're doing like who's using it and who's not.
41:04 >> Yeah, you could even get put targets on it like we you need to use AI like
41:08 certain percent like if you want to like sell it is probably like a little bit of a of a
41:14 build in the beginning but once you have something like that you have that
41:17 platform you can easily like redeploy it. I know that there are great like
41:21 starter kits for this already. So like for example, do you know open web? I
41:25 think it's called open web UI something like that. It it's kind of like this
41:29 chat GPT clone that you it's open source that use and you can out of the box like
41:34 this all already like you need to be a developer in order to like really like
41:37 work with this. But you you see more and more of this where you essentially have
41:40 a chat clone. You can select all the models and you can white label it and
41:44 you can just yeah deploy it on premise or completely private for a company that
41:48 could smash if you have if you have deals like that because if you can sell
41:52 that there's lot of recurring revenue because you essentially become the
41:56 vendor. I always think of like you have like rockstar offers for your agency. If
41:59 you've niched down, you've got like a real specific you sell that you like
42:02 getting you more appointments is kind of like the the classic one. Can I increase
42:06 revenue? But the a typical business strategy you have a really like
42:08 attractive thing like that that you can run ads so you can sort of talk about a
42:11 lot that gets a lot of people in. But I'm looking more and more what are these
42:14 upselles that agencies can do. It's like oh yeah you we of course we can build
42:17 you that that's going to get you a lot more a lot more lot more appointments.
42:20 Oh, but we could also like hey you probably got shadow AI usage. Why don't
42:23 we? And is you have like these softwares like the one we've just been talking
42:26 about where you're basically like a partner or a distributor of you set them
42:30 up on it. You get like a referral for your kickback from setting them up on it
42:33 and you can charge them five grand for the setup and stuff like that. I'm
42:37 really excited as this like agency market continues to kind of flourish and
42:42 and grow um what the these kinds of solutions that pop up um in the
42:45 marketplace because it's just going to allow you to expand your revenue so so
42:49 much and and a lot of the hard work's already been done by the by the SAS
42:51 platform. Of course, you always find that if you have your foot in the door,
42:55 like you said, you have a fancy like offer that you can uh like sell on the
42:59 front end. Whenever you just do a little bit of digging in any company, this just
43:04 like literally talking to people, it's so easy to find use cases like and it's
43:11 it mostly comes down to like like process processes that are just like
43:16 like not optimized. So like every like every company especially if you grow you
43:20 you have this, right? And AI is so good in that. We also as an agency have like
43:25 struggled with this at times like what do we want to focus on? And then every
43:28 time we think okay look this is kind of like a common pattern that we see. I
43:30 think this is maybe something that we can productize something around and then
43:34 another client comes in and like look something completely new. It's like oh
43:36 yeah this is also something you can do and then another client has something
43:40 completely different. So it's also like I really liked and that's also what I
43:44 like about the MCP ID right where you can kind of like standardize some kind
43:47 of an offer is standardize an offer and then you can like get a foot in the door
43:51 and like look we can like do do this build in like a couple of days we'll
43:54 give couple of MCP servers your employees are already more productive
43:58 and then they're going to run into limits and they're going to ask
44:00 questions right hey can it also do this can it also do this and now that could
44:05 like spiral off to uh bigger builds consulting uh more work >> we find it as well when you work with a
44:11 company who comes in like they come in hot and they know exactly they seem to
44:13 know exactly what they want and you're like, "Oh, well, I'm just I I'm here to
44:16 do what you tell me to a certain degree." Um, that might those those
44:20 people tend to be a lot more difficult than if you sit them down and like,
44:23 "Okay, let's let's like get to know your business first. Let's figure out like
44:27 there's way lower hanging fruit in most cases." And it's so refreshing for us
44:30 once we started doing consulting and then doing development off the back of
44:33 that. It's like, "Oh [ __ ] we were like doing the hard stuff all the time cuz we
44:36 were letting them tell us what to do." And so there's like absolute layups that
44:40 are that are hidden in most businesses that are that they're not taken
44:41 advantage of. >> That is so relatable that we we in the
44:45 beginning had that all the time. And we also had to um in the in the beginning
44:51 when uh chat GBT and this whole new thing was was new, right? We were also
44:55 excited as an agent, right? Like sure, yeah, we'll build it. Like that sounds
44:58 cool. Let's try and build it. And over time, we found that, oh, this is
45:01 actually pretty hard. And we would have projects that would stay stuck at the
45:05 proof of concept, right? So, we do like a big build and then it it kind of
45:09 worked, but it just wasn't good enough to move to the next stage. And it wasn't
45:13 really because the tech wasn't that good, but it's just like the business
45:16 case didn't make sense, right? It was actually too big, too hard of a problem.
45:20 And if you really want to solve it, like you put you need to put like another six
45:24 figures into it to like get it to a degree and then the business gets
45:27 >> and then it's still kind of a dice roll as well, you know, like
45:31 >> you never know. Yeah, you could put even more into this thing like run up the
45:35 bark up the wrong tree even further and like >> and and as an agency owner, you kind of
45:39 just like it's not even worth sure, it might be a lot of money, but it's just
45:42 not worth the stress at that point. So that so that's also that's also a skill
45:48 uh that you need to develop knowing like which are the use cases uh that are very
45:55 doable which ones are tricky and how to take the the big ones when when clients
45:59 just ask hey can you build this out for us to really say look this is a really
46:04 big problem let's start here because I know we can do that instead of like
46:08 tackling an entire big problem and automating an entire department or
46:11 something like that >> and I think what you said before about
46:14 like the MCP server being a sort of inroad. I think that's a really really
46:17 smart strategy these days is finding something general. Um I mean or I
46:21 suppose it's a specific I know every company's going to want to like get more
46:25 leads or like but in this case like I know all of them are probably wanting to
46:28 get their employees to be more productive. So if you can find kind of a
46:31 general small offer that's quite easy for you to deliver but has a has a
46:34 significant uplift for them and benefit. Hey, I'll come in and I'll set up your
46:39 team on these MCP servers and uh I'll train them on how to use it. You're
46:41 going to get a quick lift. that just gets that foot in the door and even if
46:44 you just like set and forget that and they're going to get the insights on
46:47 like oh XYZ is using it the most uh these are the kind of insuh use cases
46:51 that are coming off of it then you've got your foot in the door and from there
46:54 can come the consulting or the development or further education I
46:58 really think these these companies do need to be eased into it and that's a
47:01 really really good way to do it so uh if you're smart and you're getting started
47:04 you'll try to find um offers like that just to summarize about what we've been
47:08 through here uh no you don't need to know how to code to succeed in AI And of
47:13 course like success is is that is it outright monetary? Is it like making a
47:17 billion dollars? Is it just sort of liquiding your job? There's all these
47:20 different versions and it's such a diverse range of things that have
47:23 appeared. The market's really taking shape now. Um Dave has given us a
47:26 rundown on how if you wanted to go the the developer GI developer route, the
47:29 things you need to know. I've broken down how if you were a nontechnical
47:33 person, you can go as well. Um, but overall, if you're watching this, you
47:38 have the the chance right now to to take action on this stuff and and literally
47:43 within the space of at 6, 12, 24 months, be living a completely different life or
47:46 quit your job or whatever it is. Um, and I wholeheartedly believe that. I've seen
47:49 it hundreds of times now, as I'm sure you have, Dave, as well. So just from
47:53 from from me and I guess Dave as well like you guys are in the right place and
47:57 uh every day I'm I'm so happy to see people who have taken this opportunity
48:01 and run with it cuz um these things this this big and this much of a layup don't
48:04 come around that often. Well Dave, it's been uh great to have you on mate as
48:08 always and um hopefully have you back on and we can discuss some more uh of the
48:11 of the news as it comes out. But yeah, this has been great. So that is all for
48:14 this episode of the podcast guys. If you want to see something similar that I
48:17 really think you'd like, you can click up here to watch another one. And
48:19 remember, if you think you have a story worth telling and some valuable insight
48:22 you can share with the community, you can fill out my podcast application form
48:25 in the description below. I'd love to have a chat with you and get some
48:28 exposure for your business. Aside from that, guys, that's all for the video.
48:30 Thank you so much for watching and I'll
$

If I Wanted to Start a Career in AI in 2026, I'd Do This (Without Coding)

@LiamOttley 48:31 8 chapters
[AI agents and automation][developer tools and coding][community building and audience growth][content creation and YouTube]
// chapters
// description

📚 Join the #1 community for AI entrepreneurs and connect with 200,000+ members: https://bit.ly/3WbxgqL 📈 We help entrepreneurs, industry experts & developers build and scale their AI Agency: https://bit.ly/43uRTC3 🤝 Ready to transform your business with AI? Let's talk: https://bit.ly/47eZqGp 🚀 Apply to Join My Team: https://bit.ly/explore-roles 🎙️ Be a guest on my podcast: https://bit.ly/yt-podcast-application My Vlog/BTS Channel: https://bit.ly/LiamOttleyVlogs Connect with Dave Ebbelaar: http

now: 0:00
// tags
[AI agents and automation][developer tools and coding][community building and audience growth][content creation and YouTube]