We Built Custom Ecom Software With AI — No Developers Needed. Here's How. transcript

Sebastian Valiente is an AI nerd alongside Patrick Kadoo, my business partner who you've heard me talk with on this show a lot of times. Patrick and I have been talking about the AI tools we've been building at AJF Growth to make our work better and more consistent across the board. And Sebastian got into that conversation with Patrick via the wonderful world of DTOC Twitter, DTOC X. And so now we're having Sebastian on the show.

Sebastian's Building Growth Pod, an agency serving e-commerce clients across a bunch of different stuff that we're going to get into, but really this episode is all about how to build your own tools with AI to make your workflow better. These two dudes both know a lot of what they're talking about. I've got Patrick back with me to help guide the conversation about what we're doing on our side. Got Sebastian Valiente about what they're talk about what he's doing on his side.

And hopefully you on your side will learn a lot from this conversation. Let's get into it with Sebastian Valiente and talk about how to use AI to build your own work. [music] Sebastian, thanks for joining me and Patrick. I appreciate it, man. It's great to meet you. Great to talk to you. Uh I'm I'm grateful. I'm going to I'm going to just like I'm just going to say I'm just going to get out of the way and be like, "All right, all right, dorks.

Go ahead." You know? >> No. Well, well, let me just say I'm I'm super honored to be here. I think I mentioned to you on on DM, I've been I've been uh what's the saying? Uh longtime listener, first- time caller. I've been I've been listening to the pod for a long time. So, when you asked me to to be on, I was kind of I was kind of shocked. But, I'm super I'm super honored to be here chatting with you guys. >> Well, it happened because Patrick uh and hello, Patrick, by the way. >> Hello.

Hello. >> It happened because Patrick uh had connected with you before and he's like, "This dude's really smart. He's got a lot of a lot of stuff to say." And I, you know, just I I hadn't se seen it really or or known about it. Was it's sort of outside of my main lanes that I had been thinking about, but of course, I've been thinking about these things a lot more. And I'm joking about calling you guys uh nerds and dorks about AI stuff.

I'm I'm extremely grateful for it and extremely grateful for the ways that it has made it our work better. And so I'm really glad to have you to talk about it. Um what what's the place to start here, Patrick? I I think like I think Sebastian maybe you could just tell us a little bit about uh Growth Pod and I I would especially love to hear um what you are building at Growth Pod as a as a to set up the framing to understand a little bit more about sort of what you're trying to accomplish with you and your clients and who you're serving and then and then we can probably both talk about the ways that we're building custom AI tools to do that and and then can get into the weeds from there.

So that's that's where we can go. So why don't you start there Seb Sebastian? >> Yeah sure. Yeah. So uh so yeah, Growth Pod right now we service like you mentioned uh e-commerce uh brands we started out as a full stack growth sort of service really just working with a small handful of brands and actually so the that the small handful of brands that we started out working with we still work with to this day. Um, the thing that I had noticed as I had a kind of desire to grow was I didn't see uh a clear path to, you know, taking on more clients but still offering kind of the same quality of service that I'd been with the clients that I was working with.

And so, um, I kind of had decided, you know, towards the middle of the year in 2025 to niche down on kind of a part of what we do, but still something that we do really, really well. For us, that was performance creative. Um, and so in the last, let's call it 6 months or so, we have kind of pivoted into, we're really just at this point a kind of a creative agency. Today we're largely focused on uh I guess creative strategy and and and deep research and mixing that in with static ads.

Um and so um full full transparency, it's something that we're still kind of like actively figuring out, but right now that's kind of like the the the model that we're engaging with with brands where we are kind of coming in and uh we're sort of like a static adscreative strategy partner for the brands that we work with. Um, and you guys started talking, you and Patrick, because both of you felt like you were behind AI and both of you realized you were talking about AI or maybe Patrick just felt like you were behind.

Tell what was the what was the >> Patrick had had tweeted something about something that he had built and then I at the same time was building something internally for our team. It was basically like a uh well, it's a lot of things, but at the time it was just like a kind of like a project management tool. And I uh I like there was no one that I could tell that like cared. [laughter] So I was like, "Oh, this is cool.

People are talking about this. I was going to insert what I've done here." And I just put in the thread and I think uh I guess uh Patrick noticed and and then we kind of started chatting in the DMs after that. >> That's awesome. Um, so got a little got a little help from Elon on the algorithm, Patrick. Or were you following sub before that? Like, >> no, no, no. >> By the way, I like the ASMR you're doing here. You're drinking your drink right up next.

Sorry. We're really good. No, no, it's good. Yeah, we could probably use this audio recording for uh for for some ads or something like that, you know, for for certain things. Um, yeah. Um, no. So, okay. So, so you guys started DMing. Uh Oh. Oh, I see. and because of the because of the content. Yeah. Um so, okay. So, then what what was the initial take here? Like what did you guys start building and talking about and and help help guide me here, Patrick?

Where where where do you think the biggest things have been that you guys have been able to converge on that have helped you guys get the stuff better? >> Yeah, let me let me take over real quick and just kind of set the stage. See and I are basically both building, from what I can tell on your side, but I can speak for myself, kind of like internal agency oss. >> Yeah. like apps that that help our teams do our workflows better, faster, and with higher quality.

So, I think maybe the best place to start without, you know, instead of zooming out and, you know, painting this broad canvas, like Seb, what what I'm interested in hearing is like what is the most highest leverage thing you've been able to build and like how is how is AI basically how is Vibe coding changing how you do your work and you know um how you deliver your work to your clients? because I I can certainly speak to that answer too, but I'm curious if you'd get us started. >> And so so just to hone in on one, it was was kind of like the highest leverage that thing we've been able to build so far. >> Yeah.

Yeah. >> Yeah. So definitely would be and I'm still figuring out how to describe this because when I when I say it out loud, it sounds super simple, but it's basically so we used to run everything off of notion. We still do some things off of notion. And then that's >> um [clears throat] >> part of what happened when we pivoted into um uh into basically creative only was like I was trying to figure out okay how do we how do we essentially uh how do we how how do we do a like really really good work at like semi-cale not true scale as in like you know 50 clients but even if we were to onboard you know eight to 10 brands how do we ensure that we're actually delivering something Great.

And that kind of put me into um I guess designing systems and processes mode. Um and so eventually what ended up happening is we built out a workflow for building out our creatives that worked really really great. Um the problem with notion and maybe it's not even a real problem, but the the one gap that I saw was obviously notion is agnostic to whatever your workflow is, right? It's just a simple status and you you manage it as you go. >> And what I have found has happened in the past is like you could build like a set of rules and whatnot, enforcing them and making sure that they're followed is a lot harder and very >> Yeah.

I don't even know. I I still to this day I don't even like like without having someone that's just dedicated to making sure that that happens a certain way, it's it's it's difficult. And so the gap that I saw was like what if our workflow what if what if the project management software was exactly tailored to every step in our workflow. Um not necessarily just to enforce rules but just ultimately to make it more efficient.

And so it's like um so I'll give you an example. Um, right now a big part of our workflow is strategist briefs uh based on kind of like certain messaging and and and a theme. Let's say uh designer one will then generate imagery. Uh that gets approved and then those become individual ads with copy and then designer 2 or whoever else it is turns that into compositions. Now in notion you could obviously you know create statuses for each and that's fine. within our software.

Um, when you create a new card, the first thing that it'll ask you is, okay, enter all the details about the brief. >> Um, when then then when it gets assigned to the designer and it's his turn to, let's say, submit imagery, it is the UI is specifically designed for submitting imagery and it can't move into the next stage unless he does that. And so >> um that just as an example, but basically what it's allowed us to do is tailor the project management software directly to our workflow. >> Um which both ensures that things happen the way that we need them to happen, but also and this is what uh I've loved the most is like the feedback that I've gotten from my team has been like it has like I have to think a lot less now in my day-to-day like Yeah.

Exactly. I'm not having to figure out okay what do I need? It's just like I I I show up, I open the screen, I I I put my stuff, and then I'm off like I'm on to the next thing. Um, so that's we've built a ton of stuff, but that has still consistently been the thing that I think has like Yeah, like at least two to 3xed our output. And not just output though, it's like it's ease of it's like convenience. It's like I can tell that my team is not as like burdened, which I like.

Yeah, it's overwhelmingly my favorite part of these kinds of AI tools is it it really has surfaced for me and and this is where like yeah, like I said, I jokingly disparaged you both in the intros by calling you names, but I I actually am like so grateful for it because like what it really did is for me is it has made me realize how much of that stuff existed that I didn't even realize was there because as you take it away, it's like, oh, I don't have to like do these details in the same way I used to.

And I hate those kinds of details. Patrick will tell you like I get really really burdened by them. They really weigh me down. And it's all of these things that are not really the work. They're really not. They're just like ways to make their ways to make the work >> get to the next person or whatever, but they're not really the work. They're not the most interesting or fun parts or anything like that. And in those respects, AI, that's where like I have found this kind of stuff that Patrick has built for us to be so freeing and so just delightful.

Like it it creates more joy in my life, you know? Exactly. Um Yeah. I don't know Patrick what you want to say about that but >> yeah you said exactly what I was thinking which it makes more joy in the work so instead of think thinking about >> uh how do I format this table and what was the name of that ad I'm trying to make again you know it's to your point Seb it's like notion was a great tool for that but it's so workflow agnostic that you know when you can build something exactly to what you want it to be like it it it's a gamecher um and and the beautiful thing that you we haven't talked about yet is at least for Hey, I haven't touched a line of code.

Um, I know you you have a little bit more of a software background and maybe you've touched a little more code, but you know, for the most part, you know, both of us are vibe coding. Um, and it's just amazing to be able to turn what used to be such a a cumbersome, painful process into a joyful one where the people like we have, you know, you have artists on your team and I have artists on my team and they didn't join, >> you know, growth pod or AJF to do, you know, to make to write tables and like figure out notion.

They joined to like make art. And so we want to free them up to do their work. And so I think I'll say one last thing and then I'll I'll turn it over. I think the reason I want to talk about this topic is like I think this is the most interesting topic for vibe coding is building internal apps that are customtailored to what you do and take your knowledge and put it into like an actual software. Um, you know, nobody's gonna code a Netswuite or like a Clavio, you know, like like those tools aren't going away.

Although they're probably going to get, you know, they're margin, you know, >> those tools are definitely going to become smaller and smaller, but like the ability to create these internal tools, I think, is huge. So, I don't know. It probably it I guess next question, Seb, like it probably sounds boring to you. um because it sounds boring to me a little bit to share, but whoever's listening to this is like probably really interested in like the specific examples.

So, I can give some u but I'm not the guest, you are Seb, so I don't want to do all the talking. Uh you just gave a great example. Does anything else come to mind just like really specifically what you're building and also like how you may be building? So, there's two pieces to the AI here. One piece is I use AI to build the tool. >> But then there's another piece which is I use AI inside the tool to generate for us.

We generate copy inside the tool with AI. You maybe you generate images. I I don't know if you told me you do that. I know other people I've talked to do. >> Yeah. So right now so actually funny enough right now for anything that's like generating assets whether that's image or copy. kind of already had workflows outside like that like within let's say cloud >> like the UI is like using cloud projects >> we use a lot of Flora if you guys have heard of that it's like a canvas UI so we're using that for so like a lot of >> I don't know Flora tell me let me speak for the for the people who don't know anything >> so Flora I think on if you look it up it's like flora it's just like a canvas tool think of like a Figma and you like plop in a square and you can select an image model give it the prompt u it's just It's just a more creatively intuitive way of interacting with image models versus like a chat window if that makes sense. >> Yeah. >> Um >> yeah.

So, so we're not I I haven't attempted and I probably won't for some time just because the stuff that we're doing now works um to like replace stuff like that. But um I for example like stuff that we are using AI for in the actual platform is like um two a couple of things. So, let's say uh the designer that's building out the composition for an ad submits it. That goes through uh various AI checks to check for things like uh typos.

Um is there enough contrast? Is there anything that's like super off-brand that the client isn't going to like? Things like that that otherwise some would have had to do a manual QA. And we still do manual QA, but having that extra layer there that no one else has to do is nice. But that's that's I mean nothing else that we're using like actual AI for a lot of this stuff is like really kind of like small and menial things that like you know it's just like no one's going to want to do it anyways type of thing. >> Yeah. >> Um >> that's a good framing. >> Yeah. >> Make it do all the things nobody wants to do and then give the give give the artist the freedom to kind of use the tools that they want to do the work. >> Yeah.

Exactly. If you're building an e-commerce brand, you need to service your customers when they have questions. And the best customer service help desk software to use for that is Rich Panel. Rich Panel, first of all, right off the bat, guarantees you 30% savings if you switch from Gorgeous or Zenesk over to Rich Panel. So, right away, there is a strong reason to consider them. Secondly, you should know right away also that on average, Rich Panel sees a 30% reduction in customer service tickets for people who switch from Gorgeous Oendes to Rich Panel.

So that is a couple of really big claims right away. You're paying less. You're seeing a serious ticket reduction. That means you have to put less agent hours towards things. And all of that is possible because Rich Panel was built from the ground up AI first. And AI of course can make a huge impact of all places in your business on your customer service situation. And it even has uh the ability to do automated AI interactions on your organic and paid social so that rich panels AI agent can go and respond and interact with customers for you [music] which is awesome.

If you've been in the meta ads game for any amount of time you know that's been a huge pain for forever. Customers are asking questions on your ads. They have thoughts on your ads and it's just [music] really really hard to stay up on top of those. There's never been a good way to do it. Rich Panel's AI agent can do that for you which is a huge advantage. Also again reduces the amount of human hours you need to put towards that kind of a problem.

Um, Rich Panel is built to work with you if you are at $0 and just launching your brand all the way up through nine figure plus brands. Like Ridge is using Rich Panel for their customer service help desk software. So, uh, if you think it's just for small brands or something like that, it's not. It is. It can scale with you as your entire business grows. Get on a call with them. They promise a twoe transition from other software over to them.

They know you don't have time to waste getting agents up to speed and all those kinds of things. So, go to richpanel.com. Get a conversation going today. Tell them I sent you richpanel.com. I hear a couple of things happening in this conversation and it it's definitely hits some of the this is slightly less specific. I I actually would love for everybody to maybe show some examples of some of these things or or something.

I hear I hear when you're talking Sebastian a mix of here's the way we're building custom tools and then here's also the the sort of publicly available tools we're relying on and here's actually where we're not building custom tools as well. like you said there's some kinds of work that that you're not building for and I'm and you might not have the answer here but I think almost every organization is wrestling with this right now which is like >> how much should I go super custom and build myself versus how much should I use publicly available versus how much should I actually keep AI out of this like or you know I mean like >> and I I don't know if you have any >> central principle that you're using to evaluate that or if it's just a matter of resource allocation or it's definitely something Patrick and I have wrestled with a lot Because because it it sounds like when you're describing your setup, there's a mix of all three of those things. >> Definitely.

Definitely. I think Yeah, we I think we talked about this Patrick one point like there be there are cons to building things yourself. Um and it' be interesting to chat about those. I think like the like uh for for me personally, it's it's in terms of deciding what to build and not to build, it's really a matter of like okay, first of all, is do I have a clear path in my mind to how this would get done? Yeah. >> Okay, let's say that check gets passed.

Great. How long would this take me to do? And now having built a couple of different things. I also have a point of reference not just for how long it'll take to build, but like what do I think some of the maintenance will be on that? Because the reality is just for everyone listening is like you're going to build it and then it's going to break and you're going to have to go and fix it. You're going to have to allocate time to fixing stuff. >> Um, for sure.

So >> this feels like this feels like a core danger to me of this whole thing which is like it's almost like it's so easy that you don't it's almost like when people describe buying a house where it's like oh you buy a house and that's great and what nobody thinks about is like property tax and and when your roof need when you need a new roof and like and people don't budget that and that's there's like a budgeting of time and and some of that kind of things where it's very clear that you need to >> that there's this customization but is that the main I'm sorry to cut you off.

Is that the main limiting principle in the way you're thinking about >> sort of where some of these tool stuff and start >> that and and yes that and well no yes and no. It's like to usually because the thing is if I'm going to build something again it's to me it's easy enough that I can build it and if it's easy enough that I can build it um it'll require some maintenance but not a massive amount. So it's more so just a matter of like time and prioritization because right now I'm the one that's building it.

Like no there's no one on my team that's helping me necessarily. So, it's like, okay, if I'm going to spend time building something, it has to be obviously extremely strategic and and worthy of, you know, taking my attention, focus. Um, but what's interesting is, and like this will be a good interesting contrast between me and Patrick, is like I have a like I I mean I was a software developer for like three years prior to being in marketing.

And so the things that I will consider, you know, maybe easy or difficult or or longer or shorter in terms of time to build might be different than Patrick, right? And that that's the other difficult thing to you kind of have to know your own limitations to be able to figure out. It's almost like a question that there probably isn't one sizefits-all answer. It's like it's going to depend on you ultimately and like what you have available to yourself because yeah, if you build something, you don't know how to fix it.

Uh you're you're kind of cooked. There's a very specific way I think about this and um you know when people ask me what you know how do I start doing this myself like the I I think I'm concerned a lot less about the software and the bugs and like the mess that I'm making on the back end which I'm sure if like a real software developer looked at my codebase they'd like you know have a heart attack. I'm less concerned about that now for a couple of reasons.

Um, one like the the tools are only going to get better at like debugging and, you know, maybe even doing a complete refactor, you know, at one point, like just completely rebuilding the whole app from scratch. And even if they don't, like I'll hire somebody that's way better at this and cheap enough to just go fix it all for me. And I know I know there software developers like cringing at that, but like that's just the reality. like this is going to these problems are going to get so much easier to solve that I'm not concerned about them for for the tiny little tool that I'm building and it's it's it's becoming bigger and more complicated but in terms of like real enterprise software it's nothing compared to that you know so I'm not concerned about >> I was with a uh sorry finish I don't want to cut you off >> yeah I guess the only just the other only other thought I had was like so that's a piece of it and then the other piece of like knowing what to build like that's the more interesting piece which is I don't actually have to know anything about software really.

I mean I understand the foundational underpinnings but I don't actually have to know what React is or like you know what you know what I'm actually built. I just the most important thing to know is what you touched on Seb which is like what is it I'm trying to accomplish here and what is this workflow? And so for Andrew and I we were just talking about this before the call. It's like, you know, part of us doesn't want to talk about what we're doing because it's become so easy to replicate.

Uh but the other part is like we actually think we're the, you know, some of the smartest thinkers in the space. And you can't build any of this without having a clear point of view of what it is you're trying to accomplish. >> And so for us, like the our creative briefing workflow, like that's what we mainly use it for. like I spent six months like actually honing exactly the steps that the team needs to take, who needs to take the steps, what is it the outputs are, you know, and so anyways, I'll hand it back to you, Seb, but like I think that's the real skill here because all I've done is prompt AI to to to take what I've already designed and turn it into code.

And so that that was the real hard part to be honest. >> Yeah. What what I was going to say earlier was and this is just more so just for people listening is like I I was with a a friend of mine who's a CTO at a at a a big AI company and I was showing him a bunch of the stuff that I had built and uh and his his one of his only questions to me was like how are you dealing with like uh it was like race conditions and like oh like basically issues of scale uh and I was like well we don't really have to deal with that cuz it's just us >> we got 20 people using this thing you know >> right what I have found is most of the problems, not not all, but a lot of the problems that like super, you know, serious software engineers fix are issues related to scale.

Not when an app is being used by like 10 people, even 20. Like that's nothing. You know what I mean? >> Um, and so it's like >> it's when when you're when it's not massive software, it's much more forgiving. Like it doesn't really matter if your code is quote unquote bad because it's like it's not like we're trying to ship this out to, you know, a thou thousands of users. Um, >> and if there's a That's a good thing. >> If there's a bug, your your whole company, you know, doesn't tank.

It's just, you know, you got a pissed-off creative strategist that's like, I can't do my work and then you fix it, you know. >> Yeah. Exactly. Yeah. Or you you run a workaround in the meantime. We've had to do that, by the way. It's like, oh, I was like >> for a brief like we at some point with one of our shoters wanted to like attach images and like you just like I didn't have time to add that and uh >> I was like just send in Slack in the meantime.

I don't know. the the the short way to summarize what we're saying is that the upsides are just so far beyond the downsides and the challenges. >> Yes. Yes. >> Well, and but what I I would I would say it's a it's actually a perfect illustration of the idea of the democratization of software that people talk about with AI where it's like the whole reason neither of you could have built these things before is because you would not have had the time to go actually code >> exactly >> an app that is this small.

So instead you deal with the thing you described earlier which is like the shortcomings of notion as a specific use case for a brand and you figure out you go have a conversation with Jess Bachmann about how fire teams using notion for their workflows and you learn from them and it and I I had that conversation with Jess uh a couple years ago and it was like oh this is a crazy process and I used to make I used to say this about creative people all the time which is one of the reasons why creative is is hard to replicate and bad is because people don't actually have a process to do the kind of creative learning work.

I mean, when you get down to the brass tax of how this looks in an ad account, one of the one of the critiques I've had for a long time of a lot of creative strategy is that and and like you know, my whole railing against uh creative testing budgets is that they're not actually testing anything. They're they're there is no point at which you're taking learnings from an ad account and applying them into the rest of the ad account for the vast majority of people.

Jess and Fire Team was always a counterex example to this that I would gladly reference. And the reason why was because I had seen their notion dashboards and I had seen just how much the work they had put in back in the old days to to uh turn notion into a machine of process that would allow them to actually harvest knowledge as best as you possibly could from that setup. You know, even though I still think that's flawed, like they actually seem to be able to do this in a way that most people were just kind of blowing money. >> And the thing is that was many endless countless hours of work that was hard to replicate as they grew and all these things.

And they had to share how to do and they had they custom and whatever. And now it's like, oh, you don't have to go do all of this stuff for like one process for this one thing. Instead, now it's like now you and and you don't have to have a team of developers and you don't have to have a, you know, pathway to a billion dollar outcome to hire developers, you know, you can actually just go and say, we'll just build the thing custom for ourselves because exactly what you said is true, Seb, which is like I don't need it to work for 10,000 people.

I just need it to work for my team of 20, maybe 50 someday, and that's it. And so now suddenly you can take these kinds of things and instead of shoehorning your process into notion which is built for flexibility from in the first place you know so anybody could use it for any kind of way instead you're like okay I'll just build my own little thing here and then we'll use notion less and less and less and less until until we have a custom solution.

And to me that's like that's where so much of the promise is for this kind of thing. The thing I'm still curious about is what are the actual limitations of that um in terms of like where should you not build it yourself. I'm I'm actually curious to hear you say, Sebastian, a little bit more about like, you know, what one of the things that seems obvious to me in your potential work is you've got designers uh still at the other end of this process while you're focusing especially on stills, >> right? >> Why not have the AI go build the designs? um you know and and obviously this has occurred to you but it it's actually a sort of perfect uh >> perfect little um I think test case study of like how are you thinking about where the machine stops in the human I love this question >> yeah and so so and maybe you're trying maybe you're trying to build it I don't know but like you know why not so or or what's the human doing >> yeah yeah uh yeah so many thoughts are racing through my mind I I don't want to I don't want to skip over.

I had a thought that came and this relates to this when you were talking about Jess Bachman's uh the process and the notion thing. You're saying well now you can >> you can build it out quickly and it's like you can but the one thing that you can't circumvent is like the hours that it took to think of the process >> and to ensure that it actually maps to something that is valuable. And so it's like >> it it's it's this weird kind of dichotomy where yeah, you can build stuff really quickly, >> but you can also build like useless stuff really quickly.

It's like like it being quick to build doesn't actually pertain to whether or not it's valuable. And so um but yeah, so and so another example of this is like can can I build a workflow that like you click a button and then it generates the brief and it writes the copy and it writes the images and D all entang without a human? Yeah, for sure. Will it suck? Yeah, it'll suck. Um, and I and one thing that I've like the pattern that I've seen over and over again, whether it's copywriting or uh uh visual creativity or ad compositions is like whenever you try to get the AI to to think for you or to replace you as the main thinker, the output is horrid. >> It'll output stuff, but it sucks.

It sucks so bad. I Here's actually a really good example of this. when we were building out our workflow for uh for statics, one thing that we were trying to figure out is like how do we like like basically like so we kind of break it into two parts. So like you got the visual and then the composition. So the visual is just like the underlying base image if you will without copy or anything like that. Um it's like how do we generate great images like relatively fast?

So, we initially tried, oh, what if we just take a brief and then get the AI to like prompt a bunch of different visual ideas and just feed it all at once and then great, we'll have a hundred images, right? >> Um, and that was true. We had 100 images and like 97 of them were like, I do not want to submit these to any client. And so and then so the shift that we made was okay what if uh we actually just ideulate the visuals ourselves but instead we use let's say claude or whatever other AI model to help us brainstorm and think.

Um and this was very measurable. What we saw was like when we were like leveraging our own creativity and using the AI model to like inspire versus like think for us. If you want to call it the hit rate for images that we would use went up. So instead of having to generate a hundred and have maybe three usable ones, >> we could generate 15 and like 10, we're like these are awesome. Let's move forward to the next stage.

And so yeah, there I don't really know what how to exactly pinpoint it, but there's there's a part in the whole process where it's really great to help you do grunt work or even to help you in like ideulate and to help you think of new ways of of looking at something. >> It is horrible to do the thinking entirely, I find. >> Uh so like stuff like that I'm not even going to try to build because it's like I don't I don't see a path to actually getting this done.

At least not right now. >> Does that make sense? Yeah. >> When I talk to my friends at Intelliggeems about what they are trying to accomplish for their brands, what they tell me all the time is, "We are here for excellent operators who are really dead set on driving profit throughout their whole business. That's what they care about. That's who they care about serving. And if that is you, you should be considering using Intellams to run CRO split testing on your website." And I will tell you right now that if the Intelligence team hears this ad read, they will get mad that I called it CRO and split testing because it's really not just CRO and it's really not just split testing.

It is those kinds of that kind of work that goes way beyond the basics of what you think of as traditional CRO. Intelligence can do the simple work of headline testing and design testing and all of that stuff, but it goes way way beyond that on a couple levels. At the first level, IntelligJ allows you to test the real needle movers in [music] your business for driving profit. for example, the price of your product, which has a massive outsized impact on the way people behave in terms of buying your product versus saying, "I'm going to abandon that cart." Uh, speaking of abandoned carts, they allow you to test all kinds of upsells, including down to the level of the checkout.

With Intel gems, test your free shipping threshold, which can of course make a big difference in the way people are interacting with you. your shipping charge. You can test actually full Shopify theme tests. Like the the really big stuff that's really impacting how customers are interacting with you, like I said, can be tested with Intelligj. And they can do that without using a developer, get installed and up and running really, really fast, as well as all the other little tests along the way, like design, copy, and all of that kind of stuff.

But it's not just that you can test beyond the basics, it's that you can measure beyond the basics. Intelligjs allows you to see the impact of your tests down to the level of profit, not just conversion rate, not just AOV. And they do that because they tie into your COGS data in your Shopify store and they will tell you and they will output, okay, you raised the price and you saw a lower conversion rate. Well, you also got more profit per customer or you lowered the price, you saw a higher conversion rate.

Well, guess what? You aid into your profit so much that it wasn't worth it even though you got more customers and so on. Things like that. So, you can see the impact down to the level of the thing that really matters, which is contribution margin. and ultimately profit in your business. If you're serious about growing your store profitably, you should be using Intelligjam as a core tool in your toolkit. Go to intelligjms.io and use the code ferris 20 f a r i s 20 to get 20% off your first three months.

Give it a test today. I desperately want that to be true. I I I want to 100% agree with everything you're saying. I and so far I haven't seen otherwise, you know, like right now I'm with you, you know, but the question >> and I'm happy to concede. When I see it, I'll concede. I just have not I'm not even rem like >> I've not seen it an external tool. I've not been able to crack it myself >> and maybe I'm maybe I'm just not good enough at it.

Could be a skill issue, >> but I yeah, from what from what I see it's like >> it doesn't feel like we're even close. But that's >> that's that's interesting. I mean I Yeah, like I said, I I'm I'm actually not really agreeing or disagreeing with you. And mostly I think so far that's what I what I have seen. I've seen maybe one counter example, although it was publicly posted on Twitter and so you never know whether to believe it was actually real or whether it was just like the best example out, you know, whatever.

So, uh, so anyway, it's it's really interesting. Um, Patrick, do you want to talk do you want to go into more details on some of the things that you guys are actually building? Do you want to like show it >> that I'm that I'm building? >> Either of you, both of you? Like I mean, can we can we give people more to latch on to here? I don't know how comfortable you guys are doing that. Paul. Um, >> I don't have anything kind of jenned up ready to show that's not like private client information, but I can talk through >> Oh, yeah. >> I can talk through like um >> I can talk through the basics of what I've built so far just to give the audience some ideas on where to start. >> Um, and I I'll give you this one for free.

You know where I start? So I've done let me just lay the table that I've thought a lot through through my workflows and I know we've already covered that but where I started in lovable which is the latest tool I've been using. I I experimented with cloud code at first and it it's awesome. I loved it. Um but it was a little above my technical skill in terms of just you know usability and so in certain ways. So, I moved over to Lovable and my literal literal first prompt in my new Lovable um kind of project was build me an OS for my meta agency.

I mean, that that was the prompt. You know, it's like >> that that's hilarious. [laughter] >> You know, >> that's so funny. >> To be clear, it did not oneshot anything at all, but it it built a dashboard that was like, "Oh, okay. All right, that's interesting. I can use this." And it's it's completely changed over time to where it doesn't look anything like the first prompt, but like you start somewhere and then you iterate over time.

So where it's iterated for us, it does a few things really well. Um, one, we have a brand page for all of our brands that that has a spot where we upload structured information about who the brand is, how we talk about how we talk like the brand, things you can and can't say for the brand. Like basically everything I'm building in here tries to solve a very specific problem that we've encountered over and over with a client.

So for example in the past and this is common with any agency or any creative process the client will say well you can't say that ever you know don't say X instead say Y. And it's like how do we how do we drill that into our creative strategist head? How do we drill that into our copywriter's head? Well, now there's a spot. You put it exactly in the brand dashboard and it says, "What are the things you cannot say? Fill this out." And so there's a list of everything you can't fill out.

And so what do we do with that list? We have an entire workflow about how we plan and brief our ads to our artists. So our workflows start with a weekly brand plan. Um, we just do a brain dump. The workflow turns that brains dump into structured information that can be parsed out into individual pieces of work and then we brief those ads in this this um OS um using kind of an LLM in the background to brief those ads.

Okay. So now there's a instead of having what we used to have which was a manual person go through all of our ad scripts and do a quality check to make sure that things weren't being said that shouldn't be said. Now you click a button and the AI goes, "Hey, what things can I say? Oh, you can't say those things." And it does a quality check on all of our ads before they get submitted. So like that's a perfect example of like something that used to be so painful and now it's just every time a client tells us you can't say that anymore, you put it in the in the table and you forget about it.

And not ideally you don't forget about what your client says. That's not what I mean. But like you don't have to constantly be keeping that top of mind. Again, like that's not what a growth strategist enjoys doing is like remembering what can't I say. So now you don't have to remember that anymore. You just click a button and do a quality check at the right stage in the content briefing workflow. >> So So that's one thing we do.

We've built out all of our ad formats in in great detail of like what we're building and how we're built, you know, what kind of ads we brief. Um, yeah, there's more I could say like everything from contract information for the client to like eventually I kind of I'm afraid to say this out loud because it may commit me to it, but like we use frame.io io which is like a uh you know asset kind of briefing and comment the next way to go >> like I it it works great and I don't really want to replace it but you know you start to think is the pain of building something uh worth having everything in one spot like instead of having to go to this tool to get your assets to this tool to get your copy to this it's like all there like I want you know my marketing calendar my briefs my ads my my marketing light like everything's there and so then what you can start to do is I've got a little got a little button at the bottom right of the screen that's a Prometheus chat window and you can open Prometheus and now Prometheus has all the context on everything in the app.

So anyways I could go on and on but like that's that's where we're headed and and what we're trying to build basically like a one-stop shop for all the information for everything you do and all the work you do with limitations. I'm like, we talked about limitations before. Like I I don't think I'm ever going to touch the meta API in a way that that matters significantly. Like I'm not going to build a reporting tool that, you know, outperforms, >> you know, uh, Looker Studio or, you know, it's like I I'd like to put some >> some really high level reporting like, hey, what are our top ads this week in there?

But like that sort of complication. >> So this is an interesting this is an interesting example. what to what I was saying earlier which is like it just depends on the individual person and and what you know how clearly you can see a path to building something versus not because for me >> um I don't know Andrew you could you can blurt things out right or like beep it out >> we should cuz I'm going to say like I'm I'm I don't know like I think this will be soon but like I'm pretty dead set on replacing Atrea like I don't see the need to pay the amount that I'm paying and for >> what tool is that >> atria so like creative analytics It's kind of like what we do in Adnova. >> Um, >> yeah.

And it's like for for for me personally is like I haven't found enough. There's a couple problems. Number one, I find it relatively expensive, but number two, and maybe this is a user error, but like I have not really been able because it's not tailored to the way that we look at the data, >> at least not by default. And so it's like, oh, what if I could just build something that is like it shows me exactly what I want without me having like a tool >> and I can get rid of a, you know, several hundred bill every month.

Um, >> yeah. >> Yeah. >> Yeah. The the other point of that >> I I'll I'll take that I'll take that a step further and say I don't generally use any creative analytics tools, nor have I ever. Well, >> and that's uh I I think they're kind of a distraction more broadly, but uh but that's kind of a unrelated AI, so I'll mostly leave that aside. Seb made a great point here and uh Andrew you were talking you were talking with Taylor on your last podcast with him about like >> humans just there's too much reporting there's too much information and humans don't know what to do with it so like can we create simple but meaningful reports that tell our strategists here are the exact things you need to know and here's what you need to do with that information rather than giving them access to endless reports and like endless permutations of different reports and those are helpful like >> if you're if you need to dive deep on like a really specific problem like you know why yeah I mean I used to use super metrics to figure out like why is why why is the UAE like selling so well for this product well I've got to kind of really dig into the information but like for high level creative reporting Seb when you crack that nut in your app uh just call me and I'll I'll hire you to build it in mine as well.

Yeah. Yeah. [laughter] >> Yeah. Yeah. >> Yeah. Well, I I mean, as as we're circling some of these issues, what what is is coming back to the surface to me is the question of what is what parts of the process like what what parts of the process here's here's where my mind is going. What parts of the process is the friction valuable and what parts of the process is the inherent human friction a problem? M >> and where it's a problem, you you now have the ability to mostly solve it.

And and I would I've watched this happen for AJF Growth, right? I've actually said out loud a few times, it's possible to me that AJF Growth would be dead if I didn't bring Patrick on last year. And I didn't know that at the time, but I think it's it's just because there's so many things now that can move so much faster with AI um and with the things that Patrick has built that have made it so that I like >> like we can now take some of the knowledge that we have and actually execute the work at a much much much much higher level than we could before.

And I think any of our clients would tell you that they've seen that happen over time. Um, now at the same time that has not led to all of my clients tripling in revenue, you know, and so it's there it is not it is not this thing where like it it is it just like solves the magic of everything. And because I think actually most brands are not actually like an ad away from winning, you know, there there's that kind of idea out there, but they actually aren't.

You actually have to think about ads relative to the broader process. And there's this thing about the ad creation process where I would say at that point there is valuable creative friction from smart people. You you want minds, human minds turnurning on the right kinds of problems for the brands. Somewhere out there there's this possible extinction event uh in this whole thing which is that the AI can work so much faster and longer that even if it has a much higher error rate than me, it just becomes so cheap and it's so much volume that it doesn't really matter and you just like can take enough hit swings and like but it just kind of does it all for you somewhere out there.

That's the thing that's like lingering, you know, is it is that is that it just doesn't it just doesn't matter because it can just outmuscle you, you know, but I'm not sure we're there. And there's a number of reasons I wouldn't make that bet right now. Um and and and so instead I'm coming come back to this way of thinking about this whole problem which is where in the organization are you exper uh do you have unique domain expertise that you can apply to your clients or your customers or your products or whatever in a way that creates really really valuable output and what is stopping you from doing those things and if there's something stopping you that can be automated with AI you should do it you should you should use AI for those things first and foremost and I would That's that's been the bulk of where Patrick has spent his time in our organization is like understanding how we work and what we're doing and saying, "Okay, now how do we deliver that work much faster and much better and over time continuing to pile that up is going to get you better and better and better and better and so you can produce more and produce better things, you know?" So, um, so there's that's I think the way that I'm continuing to think about this is like that is the way that you allocate resources towards it because it actually isn't as easy as you guys in some ways are making it sound like Patrick really has spent many months working on this, you know, you pointed out.

So, yeah. So, it's like Yeah. most Yeah. So anyway, that that's kind of the framework that I'm that I'm coming to is where do you want the creative friction versus where do you want to eliminate the creative friction? If you can sort of think that way, then maybe that's the place to to allocate resources. >> Yeah. What I I I love everything that you just said and uh it it excites me a lot. I think sometimes the the notion that gets thrown out there is that AI is replacing people, but I find really what it's doing is just widening the gap between like lowle performers and higher level performers.

So it's like >> Yeah. Um so and maybe again maybe some people will disagree with this but like I I actually think like copywriting is an example of this where um I it it is still I think immensely valuable to be a great copywriter because you can it goes back to the thing of the images is like you could either get like like you said you could get the AI to like spit out you know 4,000 whatever versions of a headline or you could be good enough to know like here are 10 that'll work and then use AI to inspire you to kind of iterate on them.

Um, but yeah, I don't like and so something like that's for me that's an example like I wouldn't even I'm not even going to try to build an AI or some kind of workflow process that I click a button and it just takes gives me a headline because I don't think it's yeah it's not there yet where you can just take whatever it gives you back and not kind of iterate on it and I think I and maybe this is something to say to the people that are you know figuring out where do I get involved when I go I still think that if you do nothing with let's say if you're a some kind of strategy creative strategist or version of a copywriter and you do nothing with AI and you just become like a cracked copywriter.

Yeah. Like you should adapt at some point, but I still think that you're going to be more valuable than the guy that you know is kind of mediocre but has AI. Um >> yeah. >> Yeah. >> Yeah. I think I think the problem is it's very hard to tell the difference between that and and AI may also make it so that the average copywriter now gets a much higher level like essentially essentially I wonder if what you're saying is true for the top 5 to 10% or some some number in there where that person is still going to drastically outperform AI but the thing that AI has done is taken all of the sort of middle tier uh of people who are actually not that good anyway >> and eliminated them because That's true.

Because maybe it's not a great copywriter. Maybe it's not the best copywriter. Maybe it's not. I think of my friend Lyanna Patch who I've had on the show a few times. Like Lyanna's like a thousand times better and more creative and funnier than than like most people I've ever talked to about copy. She's so good at it. And um >> and I don't think Lyanna's getting replaced, but uh because her stuff is so out of the box creative.

I mean, at least an S note. But what I wonder is if like a lot of if if like the next tier if AI is like >> can actually instead say there's a tier there's a tier below her that it can actually now you have to now be better to get above that tier. If it sort of raises that median that bar in some way to where to where a lot more people are not going to be able to actually clear that any anymore. That's that's my theory.

Patrick, you're grinning over there like like maybe you think we're both done. [laughter] >> Well, let me let me ask you, Andrew, how many copywriters do we have on our team anymore? Uh well it depends what you define a creative strategist as which we have no copyriters. >> You could merge the two. >> We have no copyriters. >> Yeah. Well, we have strategist that write some we have creative strategists that edit copy.

Um but our AI >> Yeah. functionally it's similar though. I that's kind of what I'm referring to ultimately. Yeah. No, no one is a copywriter by title anymore, but >> it's obviously still a massive part of what they're doing. Yeah. Yeah. >> But this is a good example of the creative friction thing too where it's like the like the right copywriter could uh may like this. Yeah. Like the other day I saw somebody make a graphic that was it was actually in church.

The guy the the the pastor who was preaching had a graphic that he said like I whipped this up with Chad GPT. And I and I was looking at that work and and uh you know my wife and I were talking to my wife said like he couldn't have asked somebody to like just knock that out for him instead of using Ch. Well, I don't want somebody else to do that work because it was not important that the graphic was good. There's a kind of there's a kind of graphic work that that it was functionally the same as moving something on a screen.

It didn't really require any skill. He just needed to illustrate something in in his talk, you know. >> Um, so it didn't matter how perfect it was or great it was. And it's not the kind of thing I really want to spend human hours on. And I would think about this with the copyriter thing, too, where it's like there are actually a lot of ads where it doesn't matter that it's like you're actually David Oglevie. You know, it for a lot of performance marketing on the internet, you actually just need it to say like, >> uh, we're having a sale, you know, or even just like this is a really good product, you know, and even having a sale is maybe an unfair example because there's other ones where it's like there's an unfair.

So, there's lots of other creative that actually doesn't require you to be that kind of elite tier copywriter. But I also think more brands should think more carefully about how words express their brands and that they ought to apply that kind of thinking somewhere else in their business. And maybe that elite tier copyriter can use AI to do the things that you said, Patrick, which is like make sure that people don't say the wrong thing or do say the right thing.

So there's some way in which that really elite tier copyrer like my friend Lyanna, right? Like I was talking about, could she use AI to help and I don't know how she's using it or not, but like could she use AI to help set some rules and boundaries via her creative brain across the brand that get, you know, that get into AI? You know what I'm saying? Like there's a kind of higher level of work. It's like production design versus versus like a creative that kind of thing, you know, and and so anyway, so I I just wonder if that's the way to think about this is what are the right skills for the human who's really really good and then how do they use AI to leverage them more?

So I so theoretical we got only a few minutes left. Um take let's do rapid fire. Give people a couple of things that are like the like clear ways in which they should be thinking about this right now as clear as possible. Just like what are the what are the big tips that you have for people? Go do this in your organization right now >> and go Patra. [laughter] >> I say uh open up Lovable or your Vibe Coding app of choice. >> Don't don't get prompt stage fright.

Just dump something in there and just start working. And uh maybe the one thing you do before doing that is try to have a clear idea of the first thing you want to accomplish like the first workflow you want to turn into a tool >> and that's it. just start and uh you you'll be surprised at how far you get in in a day or two. >> Yeah, I would say if you if you think about things that need to get done on a spectrum of like, you know, on one side you have really strategic things and on the other side you have really just like purely executional, it's probably better to start with things that are just purely executional, just like a a series of steps that need to get done and and kind of start on that end and and and start small.

I think plus one on the lovable thing. uh it's much more consumerfriendly especially for people that are like just have zero technical background start there and I think as you start doing it it begins to open your mind like oh I realize I can do this and I can do that. >> Yeah. >> Um but you kind of need to start somewhere for sure. >> Yeah. >> Yeah. Seb, what are uh what kind of people should hire you to do work for them?

What kind of brands? Um, yeah, if you're looking to scale up creative and uh, you know, hit new winning ads, we're we're I like I mean, we do obviously creative, but I I I I think the real leverage in what we do is still ultimately the the strategic part behind what we do. It's like um, you know, we're running weekly deep research sprints. We're trying to figure out what are the net new angles you're not tapping into, what are different ways to communicate things that you guys are already communicating the ad account.

Um, and ultimately find ads that will uh, move the needle. um then I think we'd be the the right fit. >> Can you say more about how you're doing that for people? Is that that's something we didn't even touch on? Like uh what does that process look like? What does the deep research sprint look like? What kind of information are you getting that people are not getting on their own right now? >> You know what's funny? It's like it's less I think I I I think where we on the on the research side it's less because everyone has access to all the same amounts of data and everyone can run the same prompts ultimately.

I think the thing that we can offer is like we prioritize data in ways that people either just don't have the time to or don't know how. Like very rarely do you see teams spending, you know, let's say a day or two out of the week just sifting through. Okay, here's a ton of new insights that we've kind of pulled from various sources. How are we going to prioritize stuff and how's this actually going to translate into net new creative that goes out weekly?

Um, and so yeah, that's something that we're kind of doing basically on a on a weekly basis for for all the brands we're working with. >> Yeah, it's really cool. Um, any size of brand, anything that any categories, anything else that you can help people know that you're the right right person for them? >> Yeah, I would say t I I typically categorize it by spend. And so if you're spending, you know, at least 1 to 2k a day, uh, and you kind of already have cracked that fundamental stage, but you're looking to, uh, scale up, I think that's where we're the best fit. >> All right.

Awesome. You should reach out to Seb to do that, uh, for your business. Uh, Seb, the place to do that, right, growth pod.ca. >> That's correct. Yep. >> Great. And then I've got also links to uh to Sebastian's uh, DOT to his ax account and his linked account, LinkedIn account. also in the show notes as well. Go follow him both of those places. Um, and then Sebastian, go tweet more so that we can we can do do more of this. >> Thanks.

Thanks so much for your time, man. I appreciate it. >> Absolutely. Thanks for having me on. [music] >> Big thanks to Sebastian and Patrick for that episode. I learned a lot from it and hearing these two guys talk about it. They're clearly building all the time. Also interesting to zoom out a little bit and hear some of the broader perspective about what's going on. And I've got a bunch of other great episodes coming up.

So, you should subscribe wherever you're watching or listening to this episode so that you uh don't miss out on my interview coming up with Conor McDonald from Ridge. That's going to be a fantastic conversation. I've got Shireena Bear coming up as well uh soon to talk about offers. I'm really excited about that conversation. I've got Rob Fraser from Outway Socks and a couple of solo episodes, a bunch of other stuff along the way that is all going to be excellent.

And so, you should definitely check that out. Make sure you subscribe wherever you do that and comment. If you have thoughts, questions, let me know in the comments. and I would love to interact with those there. You should go check out my sponsors, Intelligjs and Rich Panel. It's an uh the links for both of those are in the show notes. Big thanks to them for supporting the show. And of course, follow up with me at ajfgrowth.com.

All everything I'm doing is on on the website all the time. And if you want to work with me and with Patrick and see what we're doing, fill out the intake form there. Tell us a little bit about your brand and I'll get back to you, let you know what I think. You uh can also follow me on xanderj ferris. And what else? Email me podcastgrowth.com. I'd love to hear from you any thoughts you have about this episode or any other.

Thanks so much for watching, for listening.