Episode 328 - This AI Shift Feels Different… And It’s Moving FAST

[MUSIC] >> Welcome to the Smartness Amazon Solar Podcast. We really are going to emphasize the word smart on today's podcast because I think it does take intelligence to get on top of what's happened the last six months with the conversations around AI. And very specifically, we're going to talk about Claude. And I've been seeing today's guest post a lot of interesting things and talking and pushing like what's possible, building his own suite of tools or agents, everything. I think he's considered. I have with me, Brett Bohannon, he is definitely like the person to talk to for this conversation. Brett, welcome. >> Thank you Scott, some nice words. >> Appreciate it. >> So I think you've been effectively Claude Shott. That's my term that I want to end up on Urban Dictionary one day where like you've seen the power of AI. You got shot and you can't unsee it. We've all been there but it really feels like right now, like if you think of like OpenClaw, the open source agent tool that people have been using that kind of feels different. Anything that I've said so far, would you characterize it differently? Like how are you feeling about like right now? >> No, I think it's perfect. It's right now it's interesting because I reflect quite a bit. I actually just all kind of talk about like the history of all this is that I recently just reupped my chat GPT OpenAI to get into Codex to kind of have Codex and Claude code kind of talk to each other and whatnot. And I went back and looked at what I've done in the past there which was like a year or two ago or whatever and making chat GPTs or customizable GPTs. I think one of the first ones I put out there was like a bullet point checker for Amazon when they were like hey you can't do these with your bullet points, same with titles and whatnot. So it's kind of interesting to see the evolution of all that kind of happen. And I think this past six months have been faster than it has ever been in the past. >> Yeah, in the year to 18 months. >> Well let's take that GPT was end of 2022. Then let's upgrade to end of 2025. That's like three years. And I feel like the last six months is probably equal to those three years. >> And I would say OpenClaw pushed it to their light speed. >> Exactly, exactly. No, I agree with you. I don't even feel like I'm actually doing way less LinkedIn but I mean it's just ever present. And while you know you could probably argue I still think it's a valid debate like what the best models are but what is absolutely clear is that anthropic is just like center stage now. And they're having their moment in the you know the business entrepreneur, the people at the cutting edge like they are preferring it. And so actually you know there's some people that like if I talk to my family none of them have ever heard the term OpenClaw. I'm just going to say like what it is is just you install this AI machine. Usually people do it on like a separate machine and you just give access to everything and all of a sudden just comes alive as a person and can do so many different things. Access all your calendars, Slack, email, just like full context and then it's like it will do what you want it to do. And in the context of like how people think about it, it's an agent. And so well you know businesses we have agents that are like talent and labor that's what we're used to. But now we have agents that are powered by AI and they work 24/7 and they don't eat or sleep or have motivation. I was like like sorry they don't own modes, they're just always working. And I've started to create just I was just curious to see what types of categories of tools would be out there that are really like let's say AI native. You know there's some some software that are focusing on like deployed agents. You know it's like done for you tasks. And I think that's where you lean. If I'm not wrong, there are some that are just data analytics. I think of like a mixed shift data dough. Maybe some others were all they're doing is just pulling in a bunch of data and you can chat with it, understand like let's say your Amazon metrics like across the board, you know like you ask it like sales are up 10% which products or which categories, which whatever, you know, the more data you put in there, the better. So that's like an analytics AI. Then there's content creation and optimization. I look at like tools like helium 10s got that one that I was impressed by was called scalable.so. There's AI visibility. That's kind of like I've worked more in that space. It's like how visible is your brand inside of Cheshire PT? It's close to SEO. It's very close. It's different, but it's it's it's close. And then there's like MCP servers, which just allow you to grab data like I think data dive has one or did you build one? I built an adapter for data dive to plug into to basically create an MCP so you can talk to it in quad. So you can talk to what? So you can talk to all the data inside of quad. So data dive does not have an MCP, but they have an API. Okay. So I was able to build an adapter to basically bring it into quad and have the exact same API calls that you would do if you were just going to use API. As your solution, I imagine it does it like require a data dive subscription? It does. Yep. It does. It's the standard plan you have to have for their API calls. Okay. So people would plug in their credentials. I got it. Okay. Yep. Yeah. I actually it's kind of interesting because you live in a world. Everyone's talking about AI, but then there's like, I look at it in three different layers. There's kind of those people that we just talk about. And this is no spite on these people, but they're not technical and they're just chatting with the LLMs and whatnot. And then there's on the other side, there's the technical people that like to tinder and like to figure it out like myself and kind of are building things and going, oh, wow, this is crazy. Like, you know, all this kind of cool stuff. And then there's somewhere in the middle and those middle people, I like to call them like the MCP people who can install an MCP with with instructions and kind of follow that step by step. When you say install, what is that? Well, an MCP you have to install within Cloud or open AI. Okay. And so what you do from there is it allows your LLM to talk to either a hosted MCP or a local MCP. So there are hosted MCPs, so some other, you know, softwares are doing it intent-wise, I think ad labs. I don't know if theirs is hosted, but so essentially you can call that data in in the cloud and then you can talk to it and create visuals or whatever you want to do with it essentially. Now, a local one is if you're building something, so all these adapters I created into an MCP, you have to install it in the dev section of of cloud in the settings. And then essentially what you do is use plug in your API keys. So for instance, for data dive, you plug in your keys and then you can talk to it like you normally wouldn't call API. Okay. So hosted is it like the companies doing it themselves? Correct. Yeah. Okay. It's a URL callback. I did find, I don't know if this is used or ever like someone like about a year ago, eight months ago created an MCP. It's smart scout, but I mean, hell if I know if it's being used or like it's actually like functioning. But no, this is helpful for what you're talking about. Now, so what type of things are you doing with this? Like, or what like, do you think it's the biggest value for any Amazon operator? So I think, I think what it comes down to is that I talk about this often, I think about this often is everyone wants to put like a workflow into a certain bucket like, you know, this is how I do something. And generally, you can. So like listing optimization, you know, you pull your competitors, you look at this, you list it, you compare it, you know, and all that kind of good stuff. But the tools is what varies. So how do you do it? There might be a little switch in nuances and the actual steps that each individual or each agency or whatever might do it their way. And so having the ability to kind of go in and just have these tools under one roof is I think is the major unlock. So I just actually created something that's called helm and it basically bundles in all these MCPs. So I'll just give you a use case of what I did today. Working on some variations on some products. I created catalog CLI tool a couple months ago. That's evolved a little bit where essentially you update, you're putting your category listing report and then you can talk to it. It natively was on terminal, but you know, non-technical people are like what's a terminal on your computer. So I developed it so it can be an MCP. So you can do that. And I found, I'll just stay with this example real quick. I found that the item type keyword would be different than other products that I wanted to put together. And I was like, that doesn't seem right. So I had it actually call KEPA and check and see what it was actually showing up on Amazon. And there was some discrepancies. It was showing up on Amazon is a little bit different than what it was showing in the back and the category listing report. Literally, I was just doing that right before we got on this call. So that's one example. Another one is I'm reworking some advertising campaigns. And we're nitching down quite a bit. And so, you know, it's calling KEPA right now. I'm working with intent wise. So I'm building some stuff out. So I have intent wise access. It's calling intent wise. And then it's also calling data dive. So it's actually pulling a niche from data dive. And it's looking at the competitor landscape and it throws it, excelsiate saying, hey, this is what you should do. This is the campaigns you should keep based on this. This is the new campaigns. This is what we should lower it. This is your budget and all that good stuff. So it created a whole plan for me. Now, the next one is obviously, you know, create that plan, double check it, make some adjustments and how to push. That would be the ultimate, ultimate unlock. - Okay. So, for what I'm hearing, has the advantage of pulling together different data sets. One of them it's like your private data set, your catalog information is according to Amazon. And then the keep is like the public information. And like, so like right there you already have like something that is an ongoing problem. And then data dives like niche analysis. And so, yeah, then you're talking to AI with so much more context. Yeah, I mean, I think, again, this is probably like my first podcast where like, I think most people are going to be a little overwhelmed at first when you hear this. But as you know, if you start chatting with like Cloud, like, or chat to PC whatever, like you get better, you get easier, it becomes a little bit more like, first nature of like, feeling like what's possible. - Yeah. - If, you gave a really good example, but like, let's, you know, take a $2 million private label seller, what would you do with them first? I may have already asked that, but I'm just like, I want to circle around on that because I think that's like, I think a small Amazon business or an entrepreneur is like the one that's like a fast adapter. - Yeah, I think it depends on what they want to accomplish. A majority of it is, this is how I use it is I go, what am I doing daily? And this is how I've always looked at it. And how can I reduce those things that I'm doing by more than 50%, or even 75 or even 80%. - Yeah. - That's where all this started and where all it came from. And then during the process, I'm like, oh shoot, like this can do this and this can do this and I put this together and it can even go even crazier. Is it early? I don't know, probably, you know? - In a year, we'll feel like it's early, but I mean, now it's like, I don't think it's too early. - Yeah, so I mean, to answer your question, I think it really depends on what they're doing. And now that's another good point is that, if they're like, hey, I have this problem and I need to solve it, you want to solve for exactly that problem. And you use these different tools. And now if it's like, hey, I don't want to hire anyone else. I'm overloaded. Well, what are your current workflows that's overloading you? And let's plug in all these different tools and all these different schedules and automations that you can do within cloud. And it's just going to even go crazier. I mean, open AI is probably going to create some agents here pretty soon, open cloud. You can do all kinds of things. So it's really dependent on what they want to accomplish in the problem that they want to solve. You can really solve any problem nowadays. And I would honestly say if someone was like, hey, I have this problem, I'd be like, cool, let me figure it out and we can solve it for you. So skill crate, you've clobbered together a few of these agents. That's like an entry point. Someone comes in there. I see ads, skew migration, ads itself, inventory dashboard, data dives, and KEPA adapter. And what's the first step? Yeah, the first step, it does all go to GitHub. So if you're not familiar with GitHub or on that technical side, it can be a little overwhelming. But I'll give you the real cheat code, is you just take your GitHub URL, anything that's private, anything that you can see in there, plug that into cloud. And basically be like, hey, I want to install this. Or hey, I want to use this, how do I use it? And literally it will tell you. So that's going to be the best way skill crate was developed because honestly, in the beginning of all this, I was like, dude, Amazon niche and everyone in the Amazon space is slow, according to everything else, in my opinion. So it's like, I wanted to push this whole entire industry and basically be like, hey, we should open source. You see all these posts is like, I did this with cloud. I did this with cloud comment XYZ and you can get it. And it's like, no, just open source and let people get it. So that's why I did all that and open source all these tools, because it's just like, hey, let's get back to how it was back in 2017, 2018, when everyone was collaborating with each other. And it probably won't, but that's OK. But I really want to push this whole entire industry into, hey, let people build, let people have fun. And I think a lot of softwares are kind of getting there to a certain extent. Totally. I learned about one software in our space that yesterday I heard a story where two different agencies stopped using it because they were disabled to replicate that in cloud. And I think that represented $3,000 a month of software expense that they were disabled to just do on their own. And I have two reactions to that. One, I'm like, OK, we're here. We're in a new era, like buckle up. If you're one of those companies, dang it. If you're, I think there's all sorts of reasons to software. Some categories, particularly, there's just really going to change. And the second is, I mean, that's how I honestly feel. Another part is if you are building a tool for yourself, it does become something that you just have to maintain. I've built probably 30 different, maybe 40 different use cases in the last 10 years. I'm like, well, I woke up in the morning and I was either coding or I was working with a developer, and we built a use case. We started using that. The next year, we probably have to maintain it once a month. On a normal thing, it's like we had to look back like, hey, we've got to tweak this one thing. Oh, this data didn't get come through. Oh, this didn't happen. And so I'm like, I don't know if it's for everyone to start from scratch. Yeah, I just had to clean up and how the image parsing was coming in for KEPA just this morning. I was running a, so I call them snapshots. And they couldn't pull in the image data. And so I dove into it and cleaned it up and fixed it. So yeah, there is a maintenance aspect to it. And people bailing softwares-- build it themselves, there's always going to be that debate is like, do you have the time and bandwidth and resources to keep up with it, you know? So there's that. I mean, $2,000 a month for like this one agency, like that's significant, but then again, on the flip side, like if it works, but that's like what a third of an employee or 25%, sometimes like, it's worth just to pay that that like, you know, someone else is just working and getting maintained. So that's obviously a debate, but still like, I'm like, okay, so what's valuable out there and what's not? If we have these tools that can just bring in all your data and you don't need a UI, but you can still build exactly what you want, you know, where things headed, what's what's valuable in the Amazon software space? What's not? I do think this maybe even changes a little bit whether you're talking entry-level tools versus like enterprise. The enterprise ones, they're always going to want, I mean, like right now, my opinion, like they're always going to want to pay money for like problems just to get solved. Yeah. Like, so there's a space for enterprise for like the problem just to get solved, whether they're using AI or not, but at the entry level, the smaller side, absolutely like an open-source solution, it can go viral and just take off. It's going to happen. I think you're showing it. Not, you know, someone will post it on Reddit and then like 10 other people like get it, then like, you know, it just kind of like does its organic thing. Then, so yeah. Now, so I have found that there's a difference between like what's valuable in software now? Is it a workflow? Is it a UI? Not quite as much anymore. You can spin up really good-looking UIs. People take our data and just like create a fancy like UI in seconds and like it works for them. They get to create that. So then I'm like, okay, my only value was my data. And so I got to get that data better. And yeah, that's kind of the area that I'm like thinking. I'm like, okay, I guess competitive data that's hard to obtain or organize is still valuable. What's your thoughts on like what is not as valuable? Let me ask you this. Have you canceled any software? Yes, I actually have. I have canceled a software and I think you nail it on the head. It's like what is valuable? I've had a lot of conversations around this and it's interesting because you know, I've been in the space now 10 years and like I could log into a dashboard with data and look at it and be like, oh, something's wrong. And then kind of dive into that and figure it out. Now, if I can talk to an agent or a cloud with all that information and get that information and go, oh, something's wrong and then go, hey, something doesn't look right. Can you check here? I think that's where the value comes as well. Because instead of you going, okay, I got to go look here, look here, look here. You can have that process pretty much done. You know, with all these tools. Now, overall use case, like everyone talks about context and, you know, yeah, you can give cloud your data. But what's it going to say? Is it going to drafts, hallucinations? You know, how is it going to interpret it? That. So skills come into place. But it's, I always go back to, what does the user want? You know, and in this day and age, people are feeling FOMO around being in cloud or being in open AI and they're like, well, I've got to be in there and do something, you know. So if they want to be in the dashboard, that's great, but if they also want to be in cloud and talk to the same data, that's great too. And I think that's where people need to kind of be in, in general. And you can kind of see it, you know, sales force going headless and HubSpot doing all these cool, agentic stuff that you're going to have to be where people are going to be. And my bet is that people are going to be in cloud and open AI. And they're going to want everything in there. That's what I'm betting on down the road. Okay. So that, yeah, I mean, you can kind of do that now. And you're saying that like, that's going to be more common. Someone's going to pop up chat GPT and be like, okay, I like this ads agent perspective that Brett's got. I've got, I like this profitability that like, let's say sellerboard.ai, they're like, you know, you're like, I like that for their profitability. I like Brett's ad framework optimization. And like, then you're in chat GPT, those are your two skills go. People can kind of cobble their own thing. Yeah, definitely, like 100%. I mean, like, people are getting more and more like it's, so there's these like, what are these like tools that build sites like, what is it? Like, lovable or bolt. There's a handful of them. And I think like, cloud just kind of built that functionality with it. Which one? Yeah, yeah, like they, they just put it in. And so all of a sudden, like, they may just kill or adopt other industries into, you know, their chat. So you can interface and interact all of some, so that theory is that the whoever has like controls the search box is like, that's where all the value gets put to. Yeah. Interesting. I haven't thought about that. I totally think that next year, you say like this two soon, like, I think like next year, like those, those searchers will be more powerful than they are now. Yeah. So yeah. And I, you know, I can't argue against what you just said. I can't. Yeah. It's, it's wild because I threw like, I got into open claw and I was playing around with it right before prosper. And, um, I just saw it. It was, it was one of those feelings. I went out and bought a Mac Mini. It was one of those feelings that I was just like, this is, this is perfect for what I want to do. And my whole, my whole thing was, I'm a solo consultant by choice. I have, I have a book of business. I have like five different clients. Um, some of them with me for years. And I was like, man, if I could literally have employees, because I've tried to have a couple helpers, I just don't like, I don't like work and like dealing with it to be completely honest. So I was like, if I can have people doing these things for me, freeze up my time to be with my family and do the things that I really want to do and bring on more, more clients and not be burnt out and stuff like that. Now over the time, it's evolved to, oh, this can help other people and, um, you know, launch agents for, you know, customizable agents and stuff like that for, for anybody. But that, that was how this all started. And, and I was just like, I see things and I'm like, things can be done better. And, and I just, and I just roll with it and, and deploy them, chip it. Yeah. Um, uh, what is, so I, I am curious like, what's your idea that you work with a few, uh, brands or sellers? Um, are you kind of like, just like a one-man show running end to end part of the, uh, Amazon business? I mean, you're doing catalog stuff. And anytime someone's doing catalog stuff, like that tells me they could do almost do anything. I was, I, I go, I go deep into catalog stuff. Uh, I, I mean, like, okay, so I started Amazon 2016 as a private label seller, I had my own business sold it in 2018. And then, you know, started a boutique agency, sold that agency, went into a large digital marketing firm, helped develop their Amazon department and then COVID and, you know, been freelancing and consulting ever since. Back in that day, there was no, I just do ads or I just do this. It was, you do end-to-end. Um, so that's where I started was, was end-to-end. So yeah, I do end-to-end for a lot of things, but then I kind of, you know, shoveled out my niche within the Amazon space to be in the catalog because no one like doing it. You know, there's only a couple of us that do well. Yeah. Anyone that does catalog stuff, I'm like, you are the ninjas of Amazon. Yeah, it sucks, but it's it's kind of fun. Um, I like to solve problems. So it's kind of fun to deal. Um, but yeah, so that's where the, you know, my, my first thing came about was the catalog CLI audit was, you know, I had a, I had a friend that I'm working with on account and he's like, hey, can you help me, you know, audit this catalog and I was like, yeah, yeah, for sure. So I did it and I was like, oh, there's got to be a better way to do this because I know what to look for in the Excel sheep. And I was like, it was kind of fun to deal. Um, but yeah, so that's where the, you know, my, my first thing came about was the catalog CLI audit was, you know, I had a friend that I'm working with on account and he's like, hey, can you help me, you know, audit this catalog and I was be a better way. So that's how it developed was just knowing that there was a better way. And, um, and if I'm not mistaken, you built some things that like make lots of catalog updates, kind of scaling my wrong like, um, the catalog, I think you're referring to the audit thing that I have. It's your whole entire category, listen, report, and it goes through 13 different queries and it checks them against it. Okay. Okay. Yeah. And that's constantly evolving because, you know, some, some sellers have multiple skews. So, you know, attributes might be empty. And, and, and, are you scoring it at the end? You're like, hey, this is how good your catalog is. Yeah. Like, info, warning, critical, um, rufus scoring based on like three basic rufus kind of things. Um, yeah, and, um, you only get this catalog listing report. Like, isn't that like you'd like create a support request for it? You used to, which it should be on everyone's account now. Um, I haven't encountered that in a long time. Is available on the API? Um, I believe so. Um, if I don't think it comes out as clean, um, there are some softwares out there that do bring it in, but they don't have it, but I'm pretty sure it probably is. You just have to clean it up and put it together. Okay. Um, I'll check that out. I'm interested only like, uh, that's cool. I like auditing stuff and, um, giving things scores of like, you know, you're, you get to be, you get to see, um, well, yeah, I mean, a couple of people have commercialized the, the, the, what I, what I created. So, okay, it's MIT license. So you can check it out and get up and check it out. All right. I mean, I'm looking at it right now. But, um, I'm, uh, I'm already pulled in two minute or two ways. I don't want more. Like I said, I don't want more tools to maintain. That's we're going to improve existing, generally speaking. Um, well, um, so again, I think is, if someone asks questions, has questions, AI clawed right now, you're definitely a good resource. What's a good way for them to get a hold of you? Um, yeah, great question. I think, um, either just vortex.com, v-o-a-r-t-e-x or skill crate.com are.dev. Sorry, skill crate.dev. Um, you can get a hold of me both those ways linked in X, um, well, totally worth following at least for like, uh, what you're thinking about, what you're talking about, what you're building. Um, and, uh, I think you're doing this in a different way than most people. Um, you know, most people want to build a tool to sell. Yeah. Yeah. Like I just, I just launched something that I've been working on. Um, it was, it's called Helm. Um, check it out. HelmCrate.com. Um, I actually made that website with clawed design over the weekend. Um, it took me one evening, basically. But um, it's essentially what we talked about. It's, it's layering on, um, it's having MCPs on one master MCP, essentially. Um, um, because what happens is you're pulling all these different ones and it creates blow. Um, so having it go faster, um, always helps. But, um, yeah. So I'm developing that as we currently speak and, um, just kind of put it out there and into the world. Um, yeah, it's fun to build. Okay. Um, cool. Well, hey, thanks so much. I, I think, um, just talking with you, I had to like think through a few things that I frankly hadn't thought about. Um, sorry if it was just my thoughts in the sky, but I'm honestly impressed with what you're working on. So thanks for coming and sharing some of, uh, uh, your knowledge and honestly excited to see what happens at the space. We're in for a shakeup. Yeah. Absolutely. I'm excited for it. Thanks for having me on. Okay. All right. Thanks, everyone. We'll wrap up there and, uh, that's the pod.