TEA 239: How the Top Ecom Brands Are Using AI (Inside Smart Marketer's Private Event)

0:00We just got back from Smart Marketer's private mentorship event, and last year we went to Smart Marketer Live, and the conversation was mostly around acquisition, acquisition channels, creatives, all of that, with probably one session being about AI, but this time it was very different. A majority of the sessions, actually, this time around, were about AI, and in this episode we're going to talk about the three major AI lessons that were talked about at the event and how some of the largest e-commerce brands in the world are using AI.

0:32I'm really excited about this one because we have been getting into AI a lot ourselves for the e-commerce alley and have been brainstorming and thinking and trying to figure out how are we going to start to use AI, and then going to this event not only was kind of validating that we're moving in the right direction, but also gave us a huge jumping point to be like, okay, we can take what some of these people have built and amplify them ourselves and grow even deeper on AI. It was a lot of fun, and by the way, I am sick, so sorry to anybody.

1:08I will try to sniffle out of the way of the microphone. It was validating but eye-opening to how we can actually begin to use things, and I kind of feel like, you know, I wonder, is this what the founding fathers felt like when they were, or when they were beginning to explore the West? It's like this whole new world, and it's on the Oregon Trail, right? And we're trying to get out West, and there's like all kinds of things that are killing everybody along the way. I think probably like the start of the internet would be a better analogy.

1:40I feel like it's like new frontiers, and there's just such a degree of excitement, and at the same time, there's also this degree of overwhelm, and before we get into this, because if you are like, hey, AI is like, I hate saying the word AI, by the way, because I feel like it's such a buzzword, and it really is, right? But that's the way that you have to describe it, because that's what it really is, artificial intelligence. And when we get into this, as we're discussing this, we did an episode last week, I believe,

2:10on us abandoning ChatGPT for Claude. We kind of left it behind. We have moved to Claude for a lot of reasons, and we are still glad that we did that. So if you haven't listened, go listen to that episode or watch that episode. But as we get into this, I would just say this. If you feel like we're going through, don't feel like you're behind. Everybody feels like they're behind. And I think, so at the event, Molly, and I don't know who said this. Maybe Molly said, no, maybe Molly did it. It may have been Ezra at the beginning. You know what?

2:40Molly actually said this, I think, maybe on one of our private things that she did for us. Okay. She ended up saying, don't feel like you're behind, because everybody feels like they're behind. I feel like I'm behind, right? And then I'm on a session with clients, and I'm showing them some basic things the other day, and people are like, what in the world? You know, I'm like, oh, I thought I was behind. But I say this because, and not to talk down on our clients, by the way, don't feel like you're behind. Just control the inputs of what you can actually do. And that is just allocating time to learn and to grow in that area of your business. And so I would just say that, don't feel like you're super behind as we're going through

3:12this stuff. We're still working through a lot of this as well. So with that said, Dylan, I'll kind of hand it back to you. Yeah. Yeah. Another thing that came up a lot during the event was people not knowing how to use AI. And here's the thing. AI is technical. We can say all we want that AI is not technical, but it is. And a lot of the common theme that showed up at the event very frequently was, if you don't know how to do something, just ask the AI. If you have an idea, and we'll talk about this a little bit later in the podcast, but

3:45how to automate recurring tasks, you have a lot of recurring tasks that you probably do in your business that could be automated by AI. Yeah. And real fast, just so you know, this event was hosted by Molly Pittman, Ezra Firestone and the Smart Marketer team. And the group is called Mentor Table. It's like their mentorship program. And so we've become very good friends with Molly and a lot of their team. And, and so if you've ever not attended one of their events, you should consider attending one of their events. They're really amazing human beings and they deliver insane events.

4:18Yeah. Smart Marketer live is in October. So yeah, definitely recommend that one. Definitely recommend. So what I'm going to do for this episode, I, I wrote down, there were so much more that happened at this event that unfortunately I won't have time to cover, but I've just brought the three most important lessons that we got around AI and I'm just going to share them with you. These were like 45 minute plus sessions that I'm going to boil down to like six or seven minutes that will just give you a high level, something that you can, here's the thing that I found.

4:48The hardest part about AI is knowing what to ask, knowing what to do, knowing how to do it. So this is going to give you the inspiration to go to Claude or ChatGPT or Gemini and ask it, how do I do this? Or what can I do? Or something like that. Or it gives you the tools to, you know, go and figure this out and then figuring it out is actually the easy part. So the first thing that I want to talk about is an AI tool stack that was shared by Kurt Elster. He's the host of the unofficial Shopify podcast. He shared his top 10 tools that he's currently using with his team.

5:21And this is just super valuable because AI tools are like the difference between a good tool and a bad tool is very stark. And we talked about this on our podcast last week, ChatGPT versus Claude, how switching to Claude has changed how we're able to use AI and the difference between these tools will do that as well. So I'm just going to go down through the top 10 list and give you his top 10 tools. This is literally just go look into these, research them, maybe see how you can plug them

5:51into your workflow. So number one is Claude Cowork. This one is probably like the number one top, top one. It's a desktop tool. It's non-developer compared to the number number two, which is Claude Code. But Claude Cowork is non-developer AI automation. And you can work inside of folders, which Kurt was like, which sounds really uninteresting because it is. But it's also super valuable because it can output into a folder. You can save files. It can reference to a folder, all of these different things inside of Claude Cowork.

6:25That's really valuable, being able to work inside of folders. And then Claude Code is the number two, which is a little bit more technical. It uses a command line and everything. We have not gotten into Claude Code too much, but I think you're saying you're going to get into it tomorrow. Yeah, yeah, yeah. So I'm excited to play around with it a little bit. But Claude Cowork and Claude Code are like the top two. If you're not using those, you really should try to find a way to. Number three was kind of funny to me. Number three was Grok, which is XAI's AI.

6:56He said he uses this for research and real-time data because it can go in search X, which gives it more real-time data than most AIs can get. And he said this is like his backup. And actually, he said it was his backup to his backup, which is ChatGPT. But he actually finds himself going to Grok for like specific things. And even, you know, I was sitting at a table with him and he like pulled up Grok and I could see he was doing some things on Grok. So it may be a little more than just his backup, but for research and image and some fun stuff

7:26that he was doing. Number four is Adobe Firefly and Nano Banana. These are image generation tools. One thing that's kind of unique about Adobe Firefly is that it is commercially safe. So if you have any concerns about Gemini producing things that could get you in trouble, because there is a possibility of that happening. And actually, we've just seen that happen. I forget someone someone generated an image of did you see this Frank Kern? Yes, he generated an image of some girl and the girl looked very similar to some other

7:56girl in the space and that girl was mad and it was a whole thing. But Adobe Firefly is commercially safe, licensed to be used commercially, whereas Gemini is not. But Adobe Firefly and Nano Banana are great image generation tools. Number five is Whisper Transcribe. This is basically just audio to text. It's just transcription and it's just really fast. It's actually powered by OpenAI, but it's just really fast. So if you have videos that you're wanting to transcribe really quickly, you can throw it in here.

8:26It can manage like large files and everything and it can do a transcription pretty quickly for you. Also, so does Gemini. Have you ever used that yet? I haven't used Gemini. Well, no, I haven't. I'll just grab like audio or video files and I'll just drop it in there. It'll just transcribe the whole thing. Yeah, it's pretty cool. Well, we'll actually talk about that in the next segment where we get into like when they talk about their AI creative strategy, which is really cool. But they use Gemini a lot for that. The smart marketer team actually talked about that. Number six is Whisper Flow, which Josh is a huge fan of.

8:57And Whisper Flow is like the best talk to text that you can get. But I actually heard Rob doing it. He's like, insert a smiley face emoji and put in a smiley face emoji and it does paragraphs. You don't even have to say insert smiley face emoji. You just say smiley face and it puts an emoji. An actual emoji, not like a little colon. Yeah. You put say heart and it puts a heart. That's cool. Yeah. You just say the emoji and it does it. Yeah. So Whisper Flow we've been using in the office. It's just better transcription. There is a link in the description. Really fast. Yeah. Number seven is Adobe Enhance.

9:29This is if you're if you're maybe doing videos or audio something, you can run your audio through Adobe Enhance and it just improves the quality. I actually have access to this in Adobe Premiere and I've played around with this a little bit and it's really good. But you just have to be careful if you are going to use Adobe Enhance, maybe for like UGC videos or something, not to crank it too hard because it's literally a slider between zero and 100. And if you put it to 100, it sounds like a robot is talking to you. It's like, you know, when Kanye uses Auto Tune intentionally, that's what it sounds like

10:03if you crank Adobe Enhance to 100. But if you keep it around like 40 to 60, it gets a it really can improve your audio quality, make it look a lot better or sound a lot better. And I use it sometimes in in videos. It's pretty cool. Number eight is Topaz. Topaz is video and photo AI. So once again, this goes toward your like website creatives or meta ads creatives or something. It can do some pretty crazy stuff. You can unblur blurry photos. You can upscale low quality photos and you can restore light from a dark photo.

10:38It can do a whole lot more than that. But this is like Photoshop, but it's like one click Photoshop. Throw it in there and it just does it. So if you have any photos that you want to use or you maybe thought were completely unusable because of the quality or it was blurry or something, you can run it through Topaz and it will it'll fix that for you. Number nine is Microsoft Clarity, which obviously this is something that a lot of people probably already use. It's he actually mentioned it's like this is insane that Microsoft Clarity is free. It is bizarre. And it's like we recommended the clients all the time.

11:10It's like one of the best heat mapping tools out there. But the one thing that he said and why he put it on his AI list is that it uses MCP to connect to Claude. So if you know what an MCP is, it's basically the new AI version of an API. And the MCP allows Clarity to connect to Claude, which then allows you to ask Claude questions about your Clarity and heat mapping. Oh, my goodness. That's amazing. I did not actually know that. So just really fast to sidebar here. So if you're not familiar, heat mapping will show the behavior of users on your website,

11:42like what they're clicking, how far down the page they're scrolling, where their mouse is hovering or their finger is hovering on their phone. So it gives you a lot of data like that. But you still have to interpret the data. So the fact that you could feed that into Claude and then you could probably build a CRO specialist skill inside of Claude that could then analyze that. Holy cow. That's amazing. You could probably say that Claude co-worked did like do that on a regular basis to analyze behavior and provide suggestions on ways to improve the landing page or the copy or whatever. Yeah. Wow. And that's why he said that is like on even made this list. We need to tell Chris and Robert that.

12:12Yeah, that's incredible. Yeah, that's pretty cool. And the number 10 on his top 10 AI tools list is a live support chat. We recommend Gorgeous for this. He gave a few that he recommend as well, but we recommend Gorgeous. It's just an AI powered customer chat that lives on your website. There's a link in the description for that as well, that it just speeds up customer support. And he also mentioned that people are more likely to buy when they get to talk to someone before they buy. So that was that. Now, that was his top 10 list. And like I said, these were like 45 minute sessions that I'm trying to boil down into

12:46six minutes. And I think I've already gone over time on this one. So we'll see how the rest of this goes. But I want to share a few more things that Kurt shared that I think is really, really great for just foundational AI use. So one thing that he mentioned is AI slop, you know, AI slop is a buzzword that people use. And he brought up a good point where AI slop, I mean, there's a lot of AI stuff out there, but not all of it is AI slop because AI slop is things you can tell are AI.

13:21So that triggers your brain to say this is AI slop. Whereas you can actually get outputs from AI that's not AI slop that looks good. But he said the reason that people are generating so much AI slop is because they don't know how to prompt the AI to not generate slop. So he shared a really, really simple framework that I think was really, really great for building a prompt that will output better outputs. So it's three, I don't know, I guess, hierarchies of the prompt.

13:54The first one is the role. You have to establish the role for the AI, whether it's a thinker or a doer. So a thinker would be someone who maybe helps you create strategy or helps you write copy or helps you think of like higher level things. And then the doer is someone, it's an AI that just does the thing. So for instance, Josh was just mentioning, you could have something automatically go into Microsoft Clarity and analyze and give you a recommendation. That would be a doer.

14:25So your role is to go to Microsoft Clarity and pull all the data and analyze it. So that would be the role. Well, the role is, is the, like you're the CRO expert, kind of like what you said. Then the task is what it does. So task, what it does, which would then be like, like I just said, go to Microsoft Clarity, pull the data and analyze it. And then format is what it outputs it to. So does it output it just in a massive block of text that's super unreadable and terrible?

14:58Does it output it to like Google Sheet? Does it output it to a website? Does it output it to whatever? So if you are writing a prompt and you think role, is it a thinker or a doer, task, what it does, and format, how it outputs what it's doing, that will automatically just give you better AI outputs just by doing that. And then one more thing that he shared, actually, I've got two more. One more thing is he shared a resource, which I'll just put a link to in the description.

15:29It's his prompt engineer prompt. So if you struggle to create prompts for AI, this prompt that he created is really, really good at analyzing an existing prompt that you have, scoring it from A to F, fixing it, recommending fixes, and then rebuilding it, prompts or skills. I actually ran one of my skills that I built on Claude through this prompt, and it gave it like a B+, which I think is pretty good. But he gave it a B+, and then I just went and made a few different changes to it.

16:03It gave some recommendations, and now it does better outputs just by using that. So there's a link in the description for his prompt engineer prompt down below. And then the last thing, which was just really valuable, AI is really smart, but it also wants to please you. That's why often if you catch AI saying something wrong and you correct it, it's like, oh my gosh, I didn't realize that, you're so right, I should have known that. He gave three system rules that you should put into either every prompt or every, you can actually go into like Claude or ChatGPT's system rules and add this so every time it talks

16:39to you, it uses these. Number one, it's okay to say, I don't know. If you ask it to do something or you ask it for its opinion or whatever, it will not tell you I don't know at this point. But if you go and say, it's okay to say, I don't know, it'll change the way that it outputs that. Number two, push back if you disagree or see a better path. Basically, you want it to be cynical. You want it to think for itself and not just take what you're saying as the solid truth. And then number three is give direct critiques when ideas are flawed.

17:10If you give it something and based on its infinite knowledge of the world, it thinks that it can do something better. It will not tell you that unless you give it permission to. So just by adding, it's okay to say, I don't know, push back if you disagree and give direct critiques if ideas are flawed, will give you better AI outputs and allow you to actually use AI to talk to and it'll actually feel like it has a brain rather than it's just a people pleaser. So that was a Kurt session. Josh, do you have anything to add on that? No, only because I was not there.

17:41Oh, you weren't there for that one. Yeah, yeah. I had, unfortunately, I had some, I had some. Oh, that's right. Not unfortunately, but I had client coaching calls. I had to like step out and then I came back and that was one of the sessions I missed. So that was the first time you heard that stuff. Along with the last thing you're going to talk about that OmniSend had covered that was really, really good. Yes, that one's really good. All right, let's move on to AI creative strategy. This was shared by Ben Bennett, who is the creative director at Smart Marketer. And he shared how they're using AI for creative and it's probably not what you would expect. I don't think he really talked at all about generating creative through AI, which I thought

18:17was really interesting because that's instantly where people go when they think, oh, AI creative. It's generating creative with AI. He talked about it a little bit, but not really because I don't think AI is quite there to generate really, really, really good creative. Now it can, but there's still some, some flaws there, but what he talked about, and this was probably my favorite example at the whole, the whole event was he, he used AI or he used a Beaker and Honeydew example. So if you're not Muppets fans, Beaker and Honeydew, Beaker is Honeydew's assistant and Beaker is

18:52kind of an idiot. He has zero experience. He doesn't really know what he's doing. He doesn't know. He makes a lot of mistakes. He's actually kind of crazy. He's a comedic relief usually in these bits. And if the lab blows up, it's usually his fault. So Beaker is AI and we are Honeydew. We're the, we're the intelligent lab operator. We're the professor, the doctor, but Beaker, the AI is our assistant and that's how you should treat it. Treat AI as your assistant and AI will then change the amount of data that you can access, process

19:29and digest. You know, as a, as a professor or a doctor or someone working in the lab, scientists, you know, you can only look at so much data at one time. But with an assistant, you can send that assistant out to go look at all of this data and report back with the thing that it found. And that's how they're using AI for creative. So, so it's more, it's more creative. It's almost like more AI workflow within creative. Yeah. Creative analysis. Yeah. I would say the same thing. Like we don't see my observation and we were in, we're in a lot of ad accounts a week is, you

20:02know, you don't see AI outperforming most normal creative that especially video video. I don't have almost anybody that I know that that is the top performing thing. So the question is then how do I spend the most time on the highest leverage thing, which we all know, or we should know that in meta in particular with advertising, like creative is the greatest value with what you can actually do. It's not media buying or campaign structure. It's really coming down to creative. But part of the challenge is what creative do I build? And that's what I loved about Ben's methodology here is that he incorporated AI in a way that

20:35is part of the workflow and the analysis of the creative that is running and of the performance to then fuel what the next thing is that you need to build. So it's removing the, the blocks and human perspective on what's working and what's not and why. Yeah. And I think that's the thing, like, we always want to know what should I go build next? What should I film? What creative should I build? And you have to go analyze the previous stuff. And so his whole process for this is like, now he's like really advanced in this. Like if you were to watch his session, which by the way, Ben Bennett, if you're watching

21:08this, you're awesome, dude. Ben is going to be speaking at our event in October, end of September, October, which is go to attendacquire.com. It's the acquire conference. And he will probably be speaking about everything that I'm about to talk about. Yeah. But six months down the line, because right now, I mean, AI is moving so fast. So he's already, he's, he's already far above what he talked about at this event. So I can't imagine in six months what he's going to be talking about then and how they're currently using AI for whatever they're using AI for and creative at that point.

21:40Yeah. Yeah. But anyway, go check that out. I think there's like 20 something seats left. There's not a whole lot. Yeah. So, so what I'm going to walk through is once again, very, very high level, the way that they created an AI creative assistant, the beaker to the honeydew, not, they're not trying to recreate honeydew basically. So you take the creative and you run it through Gemini to get a text based description of the creative, because if you ask Claude or you ask whatever AI you want to run this through to, Hey, take this image or this video and do something with it.

22:11It's going to struggle. It's going to struggle to do that. And then also go and analyze or optimize or something with it. So they run it through Gemini first, and that creates a text based description of the creative that, I mean, it does a good job of creating that text based description, but it allows it to be AI readable better than the actual image or video itself. Then they feed that text based description through multiple prompts, not just one, multiple prompts. I think he gave an example of like 10 plus prompts that he runs that through.

22:44And then all, they all output to their analysis to a Google sheet. And something that I thought was really interesting that just kind of frames how you can think of AI as an assistant, you wouldn't necessarily want to give AI or your assistant a open book of any kind of response you can give, because then it's all going to look different and weird, and it's not going to fit your guidelines and the way you want it to work. So he gave an example of some closed variables that he used, which literally just forces AI

23:14to pick from a dropdown of creative style. Is it a video? Is it a UGC video? Is it a before and after image? Is it whatever? They have, the AI has to pick a creative style that's predetermined by them. Emotional motivator. Does it make someone feel happy? Does it make someone feel sad? Does it make someone feel angry? How does it make people feel? That's a dropdown that the AI is forced to pick from. And awareness level. We all know top of funnel, middle of funnel, bottom of funnel. It forces the AI to pick from those.

23:44Now, it does leave some open variables for the AI to create analysis off of. But, and these are long form narrative outputs with guardrails. And that's the key. You don't just want to say, figure out the ICP. It's not super, you're going to get a bad output with that. But if you say, hey, figure out the ICP based on this framework, they use the story brand framework. So, it outputs the, how the, how the creative fits into the story brand framework, persona detail, character analysis, functional versus emotional analysis.

24:20So, this all goes into a Google sheet. And he showed his Google sheet of hundreds of rows of creative that has all of this filled out. Which, once again, think about now how long it would have taken you to do something like that. Put all of your creative into an easy to digest AI readable format that shows creative style, emotional motivator, awareness level, story brand framework, persona detail, character analysis, functional and emotional analysis. It's a lot that AI is able to do very quickly. So, this is a lot.

24:51And I'm going to try to break down just a little bit different perspective to give some tangible feeling on this. For anyone that's maybe struggling to follow closed variables, open variables, here's kind of how I look at this as a way to do this. So, Gemini, by the way, the reason Gemini is so great for analyzing creative is because Gemini can actually, particularly in videos, can not only look at a video and analyze the transcript. That's what everything else does, right? Claude, you'd have to give it the transcript of the video. But it doesn't have context as to what's in the video.

25:22It doesn't know that this is a lifestyle scene happening and they're holding up the product and it's this type of a creator and whatnot. So, like, Gemini can actually analyze what's in the video content. And if you use Gemini, just a little nerdy note on this, you can give it a YouTube short link, you can give it a Google Drive link, or you can give it a YouTube link, and it will automatically be able to read stuff from a link. Or you can just upload an MP4 up to, like, I think, two gigs or something like that, and it will be able to just analyze the video within that.

25:52And so, basically, what they're doing is you could go, and this is a – actually, here's an action item for everybody. Go into your meta ad campaign or even go into, like, your organic. And I want you to take the top five creatives, like, the top five creatives. They could be statics. They could be videos. Go grab those. And then go into Gemini. And then what you could do is you could upload all of these or just take – do one at a time. But essentially, what we're trying to do is we're trying to get Gemini to, like, to evaluate what is.

26:25Like, what is this creative? Don't give me perspective on, like, well, what's the message in this creative that I need to take and build something else on? But rather, you need to figure out what's in this. What are all of the elements? Which is what Dylan's talking about. What's the creative style of this? What's the emotional motivator? What's the awareness level? Is this problem aware? Is this product aware? Is this solution aware? What is the awareness level of this? Then Dylan's talking about, like, what's the persona that might be in here? The character analysis. So, when Dylan's saying all that, it's basically analyzing the creative as it is.

26:55It's not saying, let's analyze this and do something with it now. It's saying, how do we just extract what is, and then once we've done that between, let's say you go do that for the top five creatives you have right now, then you have something that you can feed into an AI, to AI, to then help you brainstorm, what should I do next? So, what is, so we have record of that, not the bias of AI, but rather what is, and then take that, and then you go to the next phase. Yep, and then the next phase is where it all really comes together, and this is, he takes

27:29the creative data Google sheet, which is what we just talked about, with all of that information. Once again, it's just a readable database for the AI to understand every creative you ever ran, plus the ad account data. So, you just export a CSV or whatever from meta, feed that into another AI prompt. So, now we're, like, 11 AI prompts deep. This is, like, advanced AI stuff here, and it outputs a creative report. That creative report can then tell you what creatives you've over-tested. Maybe you've over-tested a certain creative style.

28:02You have too much UGC, and the more UGC you put out there, it's not adjusting anything, so you should stop doing that. Or maybe you've under-tested UGC. You've done, like, one or two UGC videos that killed it, and then you've never revisited them before. It will tell you things you've never tried. If it goes and finds you've never done a top-of-funnel awareness level ad, it'll tell you that. But, variable combos. So, this is where, if you pair an awareness level with an emotional motivator, and it always does well. Well, it will tell you, oh, when you pair top-of-funnel, when you make people happy, it

28:36does better. It'll tell you how the combos of different things work with actually impact performance. Then, variable performance by year. Once you start doing this for long enough, it'll start to tell you how things shift and move month over month, year over year, how UGC does better at this time than it does this time, or offers do better here than here. It'll start to tell you all of that. And then, you can take that report. Now, we're 12 layers deep of prompting. You take that report that AI put out, and you ask Claude about the report.

29:09And you say, hey, I want you to look at this and give me some recommendations on the next thing that I should try. Yeah. And that's basically their whole creative pipeline, boiled down into like 10 minutes. That probably took Ben a few months to make. Now, here's what I would say from that. So, like, I did this with some clients live the other day. And now, Ben, what Dylan is saying and what Ben did, Ben, you're a freaking genius, dude. Like, it's really cool to see. And when you see his screenshots and spreadsheets, I'm like, oh, I'm just nerding out in my mind. But the best way is what I just had to do.

29:40Like, literally go to Gemini and drop your top. I did this with a client. She had three videos. I dropped all three of them in there. And I said, okay, analyze what's in these videos. Spit it all out. Did it make that sound, too? Yeah, it made that sound. That's cool. So, I had to analyze the video. So, I got what is. And then I said, what ICPs do you, based on these videos, what ICPs do you believe that I'm speaking to? And then I also said, and here's a link to the brand's website. So, it had a little bit of context. So, it took that and then analyzed all of the creatives.

30:11And then it said, you're talking to it. And then boom, boom, boom, boom, boom. It spit everything out. I said, great. What kind of potential motivators and pain points are being extracted within these videos? And it was like, it did all of that. And I'm doing this live time with like 20 clients on the call. It was really cool. So, I took all of that. And I said, great. Now, I want to take. No, I didn't go through the whole what's been over-tested and under-tested. This is a simple version of it, right? In a single Gemini chat. That was it. I had dropped all three at the same time. I just did this in a batch. But it's the same principle, right? So, I took that and then I said, okay, great.

30:42Now, I need a prompt because I wanted to show them Higgs Field AI, which is another image gen, video gen platform. And I said, great. Now, help me produce a prompt that would allow me to accomplish this particular pain point for this particular ICP based on what you've helped me analyze. And it was able to like make some suggestions. Give me the prompt. I took that one of the image gen, pasted it. And it gave us some decent outcomes. I'd probably give them like seven out of 10. They were just statics. But obviously, it's just the same thing. Like you can do this without all the spreadsheets and the automations and the creative analysis

31:14and the creative briefs and all that stuff like or the creative reports. You could do that just by what I said. Yeah. As a good starting point. Yep. All right. So, the next one I'm going to talk about. I'm going to skip a little bit on this one just to. I actually have six AI lessons that I want to share with everyone that already that was like the top thing that we took away from this event or the six AI lessons I'm going to share after this this one. So, I'm going to go a little faster with this one to get to that. AEO, which is Answer Engine Optimization, is the new SEO.

31:46Kasim Aslam, I think that's how you say it, wrote the book The AEO Blueprint. And he was there and sharing about how he approaches AEO, Answer Engine Optimization. One stat that he shared, which to me was kind of scary. Is that 69% of Google queries are now zero click. Meaning, AI answers the question before anyone visits your site. I actually used this just recently. We're renovating the upstairs of our building. And I'm looking for track lighting solutions. So, I went to AI and I said, I'm looking for the best budget-friendly track lighting solution

32:22that allows for two switches. I need a circuit on both sides. So, one switch does one circuit. One switch does the other. I knew nothing about it. And I also want really unique fixtures for the track lighting. Where are the best places to buy this? Spit out two websites and I haven't gone back. Those are the two websites that I used. And that's how people are starting to search the internet. So, AEO is very, very, very, very technical. And SmartMarketer actually offers a service for this because it's so technical. But I'm just going to go through, like, how AI actually scrapes the internet and finds how it recommends things.

32:59So, the first thing that it does is, or Kasim shared, where AI pulls its citations from, which I thought was really interesting. 37% comes from just websites in general. 34% comes from blogs. 16% comes from review sites. 5% from communities. 2% from e-commerce. 2% from social media. And 1% from directories. What was really interesting, something that he shared, is that the way that AI ranks things and recommends things is first by doing the research.

33:31Which makes sense. Like, you'd want to do research before you recommend something. So, it goes and does research. But what's interesting is it does research through blogs and review sites. So, it goes and it looks at the blogs and review sites to say, okay, what is the best track lighting solution? And then it goes to your website, looks for information on your website, and recommends the website. So, the interesting thing is that you have to pass the research phase before you're actually recommended by the AI.

34:02Before your website is recommended, before your product is recommended. So, what he basically said is that your site needs to become an AI data source, not just a ranked page. So, this is where SEO doesn't necessarily matter as much. Because it's not optimizing for a human, you're optimizing for a robot. So, there's a ton of structure on the back end that you need to do to make your website readable by AI. And like I said, it is very, very technical. And you can check out the AEO blueprint on Amazon. And that will kind of walk you through his whole thing.

34:33And the reason they offer a service is just because of how technical it is. But the big thing is, and this is, Josh and I kind of are differing on thoughts on this. But I think AEO is going to become very, very, very important. Because for me, I use AEO or I use AI to find a lot of stuff. So, if something, maybe there's a better option out there for track lighting. But the AI didn't recommend it. So, I'm never going to know about it. So, optimizing your site and your products for answer engine optimization, I believe, is going to be very, very important.

35:11You know what? Good luck finding a solid meme on AI. What? Good luck finding a solid meme to use because it sucks at pulling images. Oh, does it? Well, that's the thing. That's the thing with the AI answer engine thing. You have to tell it what the image is for it to find it. It's a whole thing. So, yeah, it'll change. So, my perspective is not will it be huge and wildly important. I don't think that's – it's not if. It's when. And the question of when is when is it worth the resource to do the thing that is not trackable in an entirely black box that you don't even know if the efforts are actually getting you where you want to go.

35:52And I say this because in this whole presentation, we're sitting there and I heard over – now, again, remember, I believe this is huge. We got to – we were arguing about this like a month ago, arguing, we're debating it. And then literally later that day, Isaac's like, oh, I was doing a non-porting with a client. And I asked him where they found this and they literally said – they literally searched and clawed. And I'm like, okay, my perspective is changing, right? So, the question is like how much resource goes in? How much would this take? So, if it's a little resource of time and financial investment to make this a reality, that makes a lot of sense to me.

36:23But if it's a really high resource, because I used to blog – I've written over 500 blogs back in the day. If I was writing blogs, no freaking way. The resource of that, just like with SEO, is so high. Now, I know that you're only sourcing – building it so AI can read it. But the question is like when? Is it worth it now or is it worth it in six months or is it worth it in 12 months? And so, that's my only perspective on it because during the session, I heard the word black box like 11 times. Yeah. And I heard little correlation. We drew some correlation between this and this and – again, I believe that they do a lot of research and really good stuff.

36:58But there was a lot of like – there's no direct correlation between this and this. And there's little correlation between this and this. And it's still a black box and you can't track it and this and this. And so, my thought is like it's just not there yet. I don't know. Like in the last six months, last three months, clawed in AI has come to an abnormal level of where I'm like, oh, this is what I've been waiting for for three years. Let's go all in on this. If we tried to go all in on what we're doing right now with AI two years ago or even a year or six months ago, I would have been like, man, it would have been so much work for so little output. So, I just think it's a matter of when.

37:30That's my take. Yep. All right. And the last thing, this was like I said probably the biggest takeaway that I had from the whole event. Unfortunately, Josh was not at this one either. This is my biggest takeaway from you, through you. Because I shared it with him. I was like, Josh, this was like the best session. I'm so sad that you missed it. But this was six AI lessons that the OmniSend founder and CEO, and I know I'm going to butcher this name, Ritus. I think it's Ritus Loras. I don't know. Someone can correct me maybe.

38:00Shared his top six AI lessons and this was the most valuable session. Now, I want to preface this. He runs a software company with engineers. So, this is, but that software company is not just engineers. It's also marketing and sales and customer support and all these things. And he does not have just the engineering team working on the AI. He has everyone in the company working on the AI. So, everything that I'm about to share does not just apply to the software side, the engineering side.

38:35It applies to the business as a whole. Now, before I get into those six, I do want to share one thing that didn't come from this session but was a really good takeaway for AI as a whole. And that was Ben Bennett, once again, the smart marketer, creative director, shared that they have an AI Slack channel. It's just AI where they share findings and wins from everyone in the team. So, all the AI learnings of the whole company compounds together. I just created the AI Slack channel for the e-commerce alley yesterday and started posting a bunch of stuff in there. Because if you treat AI as your own black box of, I'm working in my own little island and not sharing it with anyone, then the rest of the team is not going to grow from that as well.

39:16So, creating an AI Slack channel or a way that your team can collaborate on AI is really, really valuable. All right, six lessons of AI from the trenches. Lesson one, AI doesn't adopt itself. It has to be founder-led. It has to be founder-led. And we saw that just recently. Josh started sharing a little bit of what he was looking at with Claude on our team meetings and doing, like, 15-minute little pieces of education on our Monday morning meetings. And all of a sudden, everybody adopted it. Everyone's starting to pick up Claude.

39:47Yeah. And that was really huge. AI doesn't adopt itself. You have to push for people to do it. And the way that we, well, the way that OmniSend did it, and now we're stealing because we thought it was such a good idea, is AI days, where you take a full working day, you cancel everything, and you just work on AI. And initially, they started with fun, making games, making apps, making things that just sounded fun to them. Because it's less scary to work on something that has no actual output than, you know, to work on something with the business.

40:22So it's a much easier way to start playing with AI. And then the knowledge and understanding that you learn from the AI will very quickly turn into tools. And those AI days after just a few weeks will start to turn into building massive things for the company. And this is something that he does every single week still to the team. And it's collaborative from what it looked like. They're all in a conference room together working on things. And he said, this is a massive upfront investment for business owners. Because now business owners are going to start looking at, okay, this is a massive salary cost to do something like this.

40:5520% of the week, one full day. Yeah. Yeah. It's a massive cost. But in the episode we did last week, we already shared that we've saved 40 hours a week just by implementing AI. That's literally hiring someone. That's crazy. And imagine what happened if we all spent one day working on AI, how much more repetitive admin tasks we could get done just by doing that. Now, I already, every Friday is my think day. So I try to spend a lot of time exploring new things and researching stuff that I document throughout the week so I don't get bogged down in the middle of something when I come across something neat.

41:30And I spend Fridays doing a little bit, not a full day, but I do a little bit every Friday. And so when Dylan first shared this to me, I'm like, freaking heck yeah, let's go. Next week, we're doing AI Day, which our first one is tomorrow. Which is tomorrow. And I'm so excited. We're catering lunch. Even remotely, we're going to video in Chris from up in Canada. And we're just going to do this thing. And I'm excited to see what happens as a result. So, hey, we'll report back and maybe we'll do a podcast called AI Days, how we wasted hours of time building apps that saved us.

42:02Yeah. Or it'll be like how AI Days destroyed our company or blew up our company in a good way. Okay. So that was lesson one. AI doesn't, what was it? AI doesn't adopt itself. Yeah. Lesson two. This was really good. And he was speaking right to me on this one. Complexity is a trap. Don't over-engineer a solution. Don't try to have a single prompt replace your whole job. First off, it's not possible. And secondly, if you try to give AI a very complex task, it will probably fail.

42:37The less complex the task is, the more likely you are to actually use it. Start really, really, really simple. And you can start to get it a little bit more advanced as you start to get more advanced with AI. But the level of advancedness that a single AI prompt can get is only so much. That's why Ben runs multiple prompts for his creative analysis. Because it's, you can't really have a single prompt do everything. Or at least not well. Lesson three.

43:07Projects versus products mindset. This was really interesting. A project, he said, is you build it or you assess what you want to build. You build it. You finish it. And you move on. That's a project. A product is you assess what you need to build. You build it. You launch it. And you use it. Don't. Because here's the thing. The first thing you make is not going to be good. The first AI prompt. The first Claude skill. The first whatever is not going to be good. So your instinct is going to be, whatever, I can just do it myself better.

43:39He says, use it, use it, use it. Because treat it like an employee. You wouldn't hire an employee and expect the first time they do a task that it's done perfectly. You would train that employee. And it may take a month, two months, three months to train that employee to do the thing. But once the employee can do the thing, and by the way, AI doesn't take that long. Now, once the employee can do the thing, then it becomes an actual usable feature, usable product that you can use. But it's a loop. It's not a build, improve, finish, move on.

44:11It's a build, use, improve, improve, use, improve, use. And it becomes a cycle of improving it, improving it, improving it. That's the only way that it's going to get better. And also, your brain actually develops. You learn new things that the AI won't learn. So you have to give the AI the new things that you've learned as a human. Number four, lesson number four. Iteration requires ownership. This is really good. Because we actually, Josh and I were talking about before this session happened, you know, do we need like an AI person?

44:43Someone that integrates AI, someone that understands AI, that's technical, that understands all the complexities and does the research and blah, blah, blah, blah, blah. And we kind of were sold. We were like, yeah, we probably should have a technical person, an AI person that's going to own AI. But then he shared something really interesting. And he's like, try and guess what team builds the best marketing AI. The marketing team. Guess what team builds the best creative analysis AI. The creative team. So you don't want a single person building a massive amounts of AI because they are not going to build it as well as the person that actually understands the thing that they're doing.

45:22And if someone is responsible for the AI, they are also responsible for collecting feedback on the AI, dispersing the AI out to the company, and maintaining it, growing it, and improving it. So having a single owner of each of these AI skills or bots or whatever you have, AI agents, you need to have a human manager that does everything. And that is what is going to allow that bot or agent to get better over time. And that person, like, the more that they go and research AI, the better they will get at AI and implementing it.

45:55And then your whole team gets, like, massively better at using this stuff. A good example of this. So if you're not familiar with Claude's skills. So skills is a great thing that, like, whenever you're working in Claude, you can push organizationally. And it's a skill that will pull into anyone's chat. So, like, we have a Josh Coffey brand voice. Like, I have a particular voice with any content that comes from me. If it's an email, if it's a caption for social media, if it's a sales page that has Josh all over it, I have a certain brand voice guide that I protect.

46:30Because I sound, I talk a certain way. I use certain formattings. I don't write massive paragraphs. I don't use a bunch of M dashes like AI wants to use M dashes. I was using them before AI before school. But anyway, so I removed them because I'm like, I'm not going to use M dashes anymore in my writing because everyone thinks it's AI when I write a lot. But anyway, so I have a brand voice guide. And then what happens is I can just take that skill in our company. We have a team Claude account. And I can go to the organization settings. And I can literally just add the skill organizationally. And so my job is to make sure that that skill in Claude is up to date.

47:00And if I ever change a way that I want my voice to sound, I could just literally say, hey, Claude, go ahead and update this skill for me. And as soon as it's updated, it applies to the entire organization. Same thing with our ideal customer profile. That's organizationally. So anything that we're building has to look through the lens of our ICP. And so Robert's in charge of that. If the ICP ever changed, Robert's our marketing director. If anything ever changes with our ideal customer profile, Robert is in charge of updating that skill. And then guess what? That automatically pushes to everybody. And so this really, really applies. I think there's a lot that this applies to.

47:31But this is very broad, right? That it needs to be owned by someone or it will degrade. But that's a practical way that we're using it inside of Claude. Yep. And it also ties into lesson six that I'm getting ready to share that I think will be maybe ease some concerns that people have about AI. So lesson four was iteration requires ownership. Lesson five is build on existing processes. Don't create new work. So he shared a story on how his marketing team would always tell him when he posts on LinkedIn, it actually gets a lot of engagement.

48:02And you should post on LinkedIn more often. And he's like, well, I guess that's a good idea. If it works and I get good response, I should post on LinkedIn more often. So he asked the marketing team to create a LinkedIn posting bot for him where all he has to do is go in and like blah, blah, blah, blah, blah, talk to the bot. It reframes it for a LinkedIn post and posts it. And he said he still doesn't post on LinkedIn because AI won't build a habit for you. It only replaces a habit you already have.

48:33So the thing that you should be trying to do is not create more work for yourself, but automate repetitive tasks. Free up the time that you are wasting on things that could be automated. And then use that time that you have now freed up to create new habits. Now, once you have new habits, you use AI to replace those habits and it becomes an automated cycle there of don't expect the AI to make a habit for you. Just use the AI to replace a habit that you already have.

49:04So that's a build on existing processes. Don't create new work. And last lesson six, a lot of the conversation in the room, a lot of Q&A and even conversations at the tables are like, is AI going to replace us? And not necessarily people were asking it, but I think speakers already knew that was kind of how people were feeling or that's, I mean, kind of how everyone's feeling. Is AI going to replace us? And I thought this was the best description of the answer or the best answer to that.

49:35It's not humans V AI. AI is not coming for our jobs. People thought this about the internet or computers. AI is not going to replace us. However, it is humans versus humans who use AI. That's where the competition really is. So if you're not, if you're not using AI, you probably will lose because it's not AI versus humans. It's humans versus humans who use AI.

50:05AI is not just going to instantly make you something, make you great at something that you don't understand. It's not going to pack and chip orders for you. That's true as well. That's true as well. But it's not going to just make you good at something. You have to be good at it. Remember, AI is just Beaker. If you just let Beaker loose on your company, the lab will probably explode. Honeydew has to give Beaker something to go work on. Beaker could not do it himself. But it doesn't mean that you should not have your own Beaker because Honeydew couldn't do what he does if it wasn't for Beaker.

50:39They're tied all the way back around to the Muppets example. I like Disney and Muppets. Rest in peace, Muppets Vision. Never even seen Muppets. It's okay. It's really funny. They just did one with Sabrina Carpenter. You should watch it. It's funny. You know. I know you won't. I'm not going to actually watch it. I know you won't. But thank you for the suggestion. Yeah, yeah, yeah. So, yeah. I mean, if it ever comes to the point of like AI versus humans, we're definitely toast. That probably is true. But at this point, it's AI with humans with AI versus humans that reject it and are like, oh, it's a bad thing.

51:10I'm not going to use this. Anyone who uses this is blah, blah, blah. Yeah. It's like, well, just literally use it to do the things. I sent an email after our last one. Our last podcast really struck a chord. Crazy amount of downloads. We sent it out an email. We got loads of responses from it. People are for it, against it, whatever. And it's just funny to think that people are like, well, you're using AI. That's like, no, like everything you do is AI. And I'm like, no, it's actually, we're doing it to replace repetitive, meaningless tasks that we have to do because humans haven't had a way to not do them so that we can serve people better.

51:42Yeah. And that's how I look at it. It's like, it's not, you don't have to choose between AI and people. You could choose people, but you could use AI to help them not have to do things they hate and do more things that they love in their zone of genius. And so that's my take. But this was really, really good. This event was fantastic. Yeah, it was. Molly, John Grimshaw, Ezra, Ben, everybody that was there. One of our friends from Meadow was there, who's incredible. And we had so many conversations. I just want to say, we just want to say thank you for the invite. We had a really wonderful time.

52:14Every time we do, we grow. And if you don't get in the room with people, I'm telling you, like, it is one of the best things you could do for your business. Like in the first night, we had ideas. Yeah. The opening dinner that we had rolled out with our team within two days that were already paid for the trip several times over. Yeah. And so just being in the room elevates you to a new level. Elevate. One of the elevates you. Talking on AI. Oh, man. I promise that was an AI prompted, guys. So, like, it takes you to this different level of, like, even what's possible. And, like, I remember one of the nights we were there, we're talking to one of the guys.

52:45He's a marketing director of Winona. Oh, yeah, yeah. Winona. And they spend, like, $250,000 a day on Meta, which I don't personally hang with anybody that does that. And so it's just really cool to, like, sit down and talk. And people that are really, like, really down to earth and able to, willing to share stuff and hop on and meet with you. And Jeremiah Allen was another guy there with Bullfrog Marketing. I think that's his name. I think so, yeah. He has a book called The Econ Flywheel. The dude's just a wealth of knowledge. And so I loved it. It was great. Yeah, it was great. If you're not in the room, you should get in a room sometime. Yeah. Well, thanks for listening, guys.

53:18We will see you in the next episode. We'll see you in the next episode.