Why I'm Hesitant About AI For Meta Ads transcript

I'm not sure AI yet can create great messaging um like a lot of people I've sort of experienced this weird like it's 80% of the way thing I'm going like I just I want it to be able to you know read a bunch of customer reviews and then spit out from me the ad that excellently communicates what customers are saying to more customers but like once it actually once the tools that I've used actually get to that level um of trying to spit it out it just is like ah not quite there and then by the time I've actually done all that work like didn't actually save any time in the midst of that another thing I said there is that I felt actually sheepish recording the episode I wanted to do it sort of as a live journal entry almost of like where I'm at with it because I think a lot of people are in this place of trying to figure out like we all see the potential but we can't quite figure out um you know how to really leverage it all the way and then the other big point I made I I think that is significant to me is that I'm actually a little nervous about the way AI cuts off the mental work involved at points with generating good ad creative like um the the comparison I made there is like it's sort of like having a math textbook when you are a kid and you have the answers in the back of the book and the point is not actually the right answer the point is to to show your work and I think the showing your work thing is how you learn to do math right it's that you have to figure out how to go from the problem to the solution um and and get the right answers and so uh and so for me what I found is that while AI was actually really useful in some ways by enforcing me to get really clear about what I did and I'm I'm thinking about other ways to use it um to help me clarify my own thinking on things when I would try to go from sort of uh value proposition to add it cut off the sort of mental churning of those middle steps and the mental churning is actually really valuable I think uh that that actually spend that time on it so those are like my big Reflections on it I'm not abandoning it at all I'm still trying to think through how to do it you responded you said you basically in some ways agreed with a lot of that but I don't know there's a few things I just threw out there which of those jumped out to you the most what do I the most wrong and what am I not thinking about correctly I don't I mean that's I guess one way to do it let me and I won't even say about wrong or right give me your reaction to that basic take on AI yeah okay let's talk about let's talk about those points kind of in order so I mean the experience that you're having is not unique to you and I think it's probably the experience that most people are having with AI right now and it's because of the limitations um in which it's used um so like the ACT ual methodology they're using um to integrate AI into their workflows and also the tools that they have available to them so let's take your review example so for example yeah I think you're right I think that I'm not sure with the exact methodology that you're using but I'm assuming you're doing some kind of getting the reviews from like your review Source copying them pasting them into claw chat GPT what one of those chat Bots um and then asking for insights um or value propositions or copy for your ads based on on those reviews um and that's probably all you could do right now with chat GPT for example with a chat bot and that's a fine way to do it but I think we can agree everyone can agree that that's not going to get you the results that you need um so something there which immediately jumps out to me is that's not the best way to use AI right now um so let's talk about a methodology that AI is really good at to get the best results so you may have seen that open AI recently released their 01 uh model which was a change to the the GPT 4 GP those um o models um and it introduced something called Chain of Thought reasoning um directly embedded into the models and what it showed which people have assumed for a while now but this kind of um quantitatively showed it is that if you introduce Chain of Thought reasoning into llms or into AI it produces far better results so what that means is that when you're doing something like review scraping and um copy generation or concept generation you have to structure it in steps that let the AI think that gives it a foundation to Think Through um and that gets the best results so for example at CTC we do something similar to this um but instead of just copying all reviews and then pasting it in in which the AI if it's doing this it's only going to like it's only going to get a summary of the reviews rather than very specific things um we actually go through every single review um under certain um a certain like rating um and positive reviews and stuff like that and we will look at the very specific reason why someone is making the purchase in a psychological framework within that within that review and we'll also create a persona for who is the reviewer in that review and we'll do that across all of the reviews and then we'll feed it into another step in the process um which is analyzing um the reasons people are buying things and what they're buying and who they are and everything like that um so that's like another step in this Chain of Thought that we're going through um and then it will kind of churn through these Concepts and then the next step once we've got the concepts is now to then go into creating copy or whatever like that so you can see how we're able to get a lot more detailed um because we're not using the chatbot we're using the API and you break it out into these foundational chain steps which actually end up producing a lot better more usable results uh when okay so how do I do that like like I mean yeah how do I do that um I mean it's okay it's going to be specific to you so you might not think about concept Generation Um the same way that CTC does for example so you're going to have a a framework in your head or framework that you use to do this manually um and it starts by defining that framework breaking it down into steps and seeing where the reviews start in that process and then unfortunately the answer which people won't like like is you need someone who knows how to code who knows who understands how to use an API and who understands how to manipulate data to create that um that system and that function for you to scrape all of that and put into a database so that's where you kind of you need to get to an extra step um to be able to get the full use out of it and that's I mean just a detour for a second that's what you are doing at CTC right is like among other things presumably is sort of um because of your ability to actually be uh I mean basically because you're a software engineer and you can think through um sort of some of those more technical details of it you're able to Leverage The Tool much better than I could do you how long do you think you are necessary for that I mean this is just obviously a guess but essentially the truth is right like what many brands should be thinking as they hear this is that I want to go work with CTC so I can get that for myself um because because you know you're building that for them but of course most of us this is the reason I'm asking most of us cannot employ a full-time software engineer to to go out to that layer so um so what I mean how long do you think we are going to be in a world where that is required to to leverage the tool that way MH well I mean two things so first of all um that is one piece of what I'm building at CTC and but the reason why they need a fulltime software engineer is a lot more than than that individual thing that workflow um CTC has we were talking about this before Taylor doesn't go 0 to one he goes 0 to 100 and we're building the 100 right now so that's kind of why they required fulltime there um so I don't think that people should dismiss that I don't think that um I think that they would be able to get help from a specialist um for certain things um but how long will they need me in that it's a good question um it's kind of impossible to tell so I think about that in two ways I would say that there's two important elements that I think that I bring to the table um that allow me to provide value to a brand or Agency for example which is that I think I understand e-commerce and I understand the data structure and I understand the intended output and then I also understand the technology and so I can bring those together um it just gets to a point where is the base model like is there going to be a GPT 5 that just understands that in um intersection so much better than I do and maybe that's the case um and maybe at that point then you can get far more without someone like me but um people are slow to adopt technology and like I'm sure you know oh for sure yeah yeah of all the brand owners you know how many of them have the time to sit down and install and adopt a new technology even if it's great even if it's amazing um there's not I think that there's like a bit of inertia in businesses where they're unable to adopt new things really quickly um and so I think that there's still value there but I mean the people who are like really honored and um the people who are willing to innovate I think that they will find a huge amount of value very quickly with even without an expert um software engineer yeah it requires a kind of cost investment that you represent for CTC that actually plenty of people could do they just I think are probably either thinking they have to do it themselves or they're not thinking about that potential way to deploy capital in the business like they just are not thinking about the idea that and I think I mean you're you're uniquely you're hard to find like you know we were joking about this a minute ago but you posted this threat about essentially like arbitraging Australian gambling like DFS I think it was or something like that right anyway like or or whatever it was like essentially daily daily lines like from different uh bookmakers and how you got banned from gambling in Australia for it like that's a pretty clever problem solving mind that is also mathy enough to get there and um and that I think is is not easy to replicate and this is so much of the question for me about about this is like what is the actual applicability of the tool at this point to me versus to CTC because I mean certainly um like you know I have five clients I'm not trying to make it a lot bigger like I'm not going to go try to find you and put it in the business I'm just not I'm trying to figure out what is the way that I could use the tool now to do that um so um so that that's I think that I think is the challenge [Music]