Amazon Sellers Special :The Future of AI and Software Innovation

Amazon Sellers Special :The Future of AI and Software Innovation Speaker 1: Hey guys, welcome back to Seller Sessions. Had a bit of a break in the last couple of weeks after the conference. Had a quick holiday and yeah, back here. I'm going to be doing a lot of content this year. Been doing a lot of changes. Brought Max back to the show. You will know who Max is from e-content and he's been on the show a few times and today We are going to follow on, for example, me and Dorian did the two hour sit down, which we broke up into a few parts. But today I've brought Max in. We're going to be talking about the impact or the possible impact on software, what we should be looking at coming up. We'll probably get it all wrong because that's how it works, right, Max? And it all happened in two weeks instead of two years. But hey, we're going to give it a spin. Max, welcome to the show. Speaker 2: Great to be back, Danny. And thanks, as always, for organizing Seller Sessions. Great fun. Speaker 1: Excellent. I'm glad you enjoyed. Right. So let's kick things off, right? So I've been open agency market. One, it was overrun anyway, too many people, not enough customers to serve. So I had to make big changes in the agency. Obviously, sellers are going to see other impact on themselves through different means outside of the Trump tax. But then, one of the possible ones that are in danger to a certain degree is going to be software. Software could be in itself and people will be able to clone certain things but there are limits. And that's kind of our discussion today, isn't it? It's like, where is that going? What are the impacts? And what should we all be looking out for? So, where do you think it's going in terms of The incumbents in terms of people that started it at the very beginning and then more people jump on it and then more people are building stuff as no code comes into the fray. Where are we Max as you sit on the cutting edge of this? Speaker 2: So I think that the software Lots of people would argue that software is going to be commoditized and that actually the base models which everybody is using now, Claude, OpenAI, whatever it would be, Are going to get so good that eventually you don't need any applications because the model can just do everything. And I guess that direction of travel is what we saw back in 2022 is, as you know, like we started the content, started off doing image generation, moved into the entire product listing. And at the time, As you know, people would be saying, oh, the images aren't good enough, they're never going to be good enough. But it's very, very clear that, you know, as soon as you saw, as soon as I saw that first stable diffusion, and I kind of knew what had come before it, because I was working at Amazon. And you know, before stable diffusion, you had those Image guesses between like a cat and a muffin. Do you remember that? Does that ring any bell? It would like, it's like almost like a capture and it would kind of guess between the cat and the muffin and everything was super grainy. And suddenly you had in 2022 a model that could actually not just identify stuff well, but generate things. And yeah, obviously it wasn't perfect and it was to some extent quite a grainy image, but it was clear the direction we're going to go in and we are going to go to a place where Open AI will do, you know, hyper focused, hyper realistic images very soon. You know, hallucinations will be, I think, by large part taken care of by the base models. So yeah, that's definitely the backdrop against which people are building the software. And then, I mean, from a provider point of view, all you can think about is like, how do you Take those base models and build something which has additional value because you have a vertical focus. So a vertical focus, you know, in our case being e-commerce and looking at connecting to APIs and, you know, adding in data, data on the audience, data on the sales data, the impressions data, all of that kind of data which a base model isn't going to have and that's how you can kind of build a thin but valuable kind of layer on top of it for people to use. Speaker 1: Yeah. Yeah. So, I mean, Brandon said this publicly. I'm in favor of it. I think Yours is slightly different, your position different and obviously these guys, you know your top three. You've got your Jungle Scout Helium 10 and of course there is the with Data Dive as well, right? And what people got to look at is what do you do when people want more or they need something to work like this and it kind of doesn't? Then you give them a pipe. You know, access to the API or framework that they can work with to do bolt-ons, because then that way they get even more value out of what they've got. They've got the original engine, shall we say. And now they've got to be able to branch off that original engine without the heavy lifting of code themselves. So I can see like a market style place appearing at some point that would work in line as and when no code becomes even more available to people that Do not know about coding or doesn't have enough information for no code. Does that make sense? Because these tools get easier and easier. That's where I kind of see it going, right? Because if you think about this, say an end user guy, I wish he did that. Then it says, it then goes and asks, it doesn't matter what software, this isn't directed to anyone. Someone will raise a support ticket. Can you do this? Can you do that? No, but it's in the roadmap or it's on the list to be put in the roadmap. And that's where it sits or where it comes eventually. Yeah. So I think with people taking more and more and as Basic use of ChatGPT, Claude etc and I'm not even including a higher level of prompt engineering. I think that gets easier as well because I said this the other day to Kevin like I'm not an AI specialist. I don't consider myself that and the way I use AI is just I talk to it like it's a guy down the pub or I don't do all structured prompts and one thing I've learned from that is I know what people are going to say. If you don't structure the prompts, you're not going to get the output that you want and da da da da da. I agree. The reason I'm saying that is there are going to be people out there just like me. Who either can't be fucked to do that and structure every single prompt and do it in such a way eventually AI catches up and understands the context of that for what you want. The reason I also say is because I want to encourage people not to be put off by it but embrace and use it more. Then you currently do even if you think you don't know how to do it much and you're not a prompt engineer. Do you understand? It's that barrier there because let's say Max does a prompt that is his first prompt or Joe does a prompt or whatever is much better than Danny's ADHD chaotic prompt speaking to his mate down the pub. I still get fucking results. Right. Better results than I would without that prompt. But then there's that next level. So I'm talking about gross roots, getting more people involved. And when the way I look at it is you asking the right questions will get you further. Then just knowing how to prompt without any creativity or thinking about what you truly want, and you're just rabbiting what other people are doing. Does that make sense? Speaker 2: Yeah, I think I listened to a podcast which featured the CPO, Chief Product Officer of OpenAI, and they had a really interesting take on this, which is they kind of see it as a bug that you have to have these people Oh, this is how you prompt. You have to write a paragraph. So they see that as a bug. That's not at all what they envision for what they're building. And it's, I think that will, you know, these, these AIs all have memory, right? So they remember the conversation back and forward. We're going to get to the point, I don't know if it's going to be one or five or 10 years. Where you won't need to be an expert in prompt engineering to get a fantastic output. You know, the AI will have memory or understand the context of all your previous conversations. When you put in a short prompt, it understands what you like and what you don't like. Because you've talked to, you know, anyone who's created images on ChatGPT will feel this, right? You spend hours going back and forth saying, not like that, do this, do that, do that. Eventually, and I'm not saying it's today, but, you know, two, three year horizon, The AI is going to have that and you mentioned the no code as well. I do believe we're going to have. Hundreds or even thousands of kind of startup companies. You don't need to be a coder anymore to start a software company. You can use Lovable or one of these other kind of builders and you can build a MVP very, very quickly, right? If you have some unique knowledge or insight and you bring that on top of an LLM, you can get paying customers without needing any coding knowledge. And you can, you know, you have the basis for which you can start to build a company. And I think it's a wonderful thing. I think in the future, I hope that we're going to have less people working at these big corporates and big legal firms, you know, and people will just be going into, I'm going to, I have a unique take on some space. I'm going to build a startup. I'm going to use all these no coding tools. I'm going to use, you know, I use Stripe to do my payments and Webflow or Wix to build my website. What used to take a team of 20 people, I can do myself with no prior knowledge pretty quickly and get a product which gives value to other people in some way. I think it's a very exciting time. Speaker 1: I think, as well, you've got to try and work to what your strengths are, right? And I think, for me, the way I use AI, let me give an example. Right now, I'm building out loads of automation and I got tired of subscribing to a software, something hot, new, shiny comes along. And then that's two weeks later. So then you take on that tool, then you've got the learning curve. Then the next one comes, that tool, learning curve. So what I do is I like to set myself frameworks. So I'm gone all in on Claude. Right. So I've narrowed everything down to three softwares and I'll explain why. So now Claude is my no code platform. And I've built a series of knowledge bases. So I use Manish for scraping. Yes, more expensive. Right. But it's the least powerful resistance and the return that you get, not that it's just free. And then what I worked out is once I scraped Claude and built a knowledge base on that, I know literally up to this point, everything about Claude. And I can ask questions like, tell me five things that a no coder wouldn't know. Or about Claude, that what Claude does, etc, etc. Right. And it lists me a load of things I had no idea it would use and most people wouldn't. That comes back to asking the right question. Yeah. Then I asked the strengths and weaknesses of Claude inside of that project, inside that knowledge base. And then I said, can we replace everything else? Deployment, databases, da da da da. And CSS libraries, you know, access to updates to React, Tailwind, da, da, da, da. And I don't know much about this stuff. Can we have a replacement to fill the gaps to use GifHub? Right. And then it's built me a long list of all of those things. So what I'm trying to do, if you imagine it, and I'll go back to my music days and the reason I say it, I want quick, fast iteration ideas. Roll out fast. Right. Start it. I don't want to build and do loads of other stuff on top. So now what I'm doing is, imagine like I was a record producer and DJ. I can engineer, I can use Neve decks, desks, SSL, the big 56 channels etc. I can use all outboard equipment and mix records to a high degree or I did 15-20 years ago. But it all starts on the laptop. It all started in Reason or it started in Logic and it was limited because back then you couldn't record vocals because it took too much processing power and storage space to record the vocals. So what did you do? You had an idea. You know the software inside out. And then you just jam and basically the basis is there. I'm using that same principle to do Insider Claude. And I know people listen and go, yeah, but you should do this and that model and 3.7, da, da, da, da. Agree. Speaker 2: Yeah. Speaker 1: But do you want to get up every day? Speaker 2: It's called vibe coding, Danny. Have you seen it? It's a big trend on LinkedIn. Speaker 1: I know. I know it is. I know it is. But then you've got to understand as well, I've worked with engineers since 2008. I've run startups. We've lost a shit ton of money. My, you know, Dr. Ellis, people know he's one of the most respected in the space, right? And obviously, I work now with the engineering team at Datadive for developing and stuff like that. So I've always worked for around engineers for a number of years. So you get to understand that People can get gassed on AI and try and use AI for absolutely everything and they forget there's Python, there's other languages, you know. There's other technology out there that does a better job than AI would or it's faster to get you there but you don't know that because you're not a world-class engineer. I totally get that. Do you understand what I mean? So I'm just trying to put myself in a box so that I can rapidly turn stuff out because that thing goes from me To the design team, to engineering. And we're all locked in on the same flow. Speaker 2: I sense this. People are always scared of stuff that they feel could replace them. So, I mean, I obviously have an engineering team. They're cautious of this stuff. Similarly, I see kind of when we work with our customers and they have, you know, graphic designers, they're cautious of this stuff. And like, you know, I have both my company and with the engineers, certainly, I now say, we want to see, obviously, when you build it, we want to build it to have structure integrity and to last and to scale. And we're going to do it properly the old way, you know, with a bit of AI coding from You know, some of the advanced coding tools, not Microsoft Pilot, I forget the name of it, but we'll use AI, but we'll do it the old way. But certainly in mock-ups, like when I review stuff internally, our team will Go on a GenSpark or Gemini and just code up the page, just what would it look like. It doesn't have to be perfect, but it'll take them, I don't know, a few hours just to get a prototype up there. And then it's a lot easier to get feedback and to make sure that everyone's aligned and to move a lot faster. So I think these tools are Oh, incredible. I know that I've seen the resistance. It's Cursor. We've got Cursor for everyone in the company. We pay for it. It's a great tool. And of course, it's not perfect, but it enables people to become 10x engineers, go a lot faster, scale up mockups, debug at scale. Cursor will write. It will kind of test it for inconsistencies. You know, a hundred ways where normally engineers will write it and they don't want to go back and like mark their own homework and check all that stuff. So there's definitely enormous advantages. Speaker 1: Yeah. And because that's the other things like I can, I had Poe, right? Andrew. Got me into that. That's where I built the suite of tools for Seller Sessions. I've got to go back and tweak those now. I've got a bit more experience. They need some fixes, but I'll do that next week. But the thing is, when you look at something like that, I wanted to build something that's lightweight in the browser that everyone would use. And I use scripts and logic via that way versus using AI because in a lot of cases it didn't necessarily need AI. Would it be better with AI? Absolutely, it could do. Will anyone use them? No, because they've got to set up an account with POE. Then they've got to burn through tokens. And there's a shift there that Meaning there is like a friction for the customer. And I said I wanted to build tools that I could give to free to the delegate. I didn't want to throw in the caveat, oh, by the way, you're going to blast through 60 to 100 bucks a month because they're not going to want to use it then and it defeats it. So I had to work out, OK, let me build lightweight tools that sits in the browser based on logic, easy to output to the engineer to put on to the website, to our site. And then it just works. You upload a .csv, you get all the results out of it because basically these are one feature tools, grunt tools that give you diagnostics, right? So they're like quick use, use those. Last thing you want is go, I'll pull that up, create a pay account, do all these different things. So it's them kind of workarounds that you look at as well. I'm not saying you shouldn't use AI. You should, I'm saying, in certain instances. What is the better outcome? Speaker 2: It enables you to go a lot faster, right? Because this is exactly how I think entrepreneurs, especially tech entrepreneurs should be thinking. You've just scaled up four or five tools for free. But the question is, do people use them? Do people find value in them? And the fact that you can use AI to do four or five You'll now have data, which is your proprietary data, what actually people are interested in. Like I'm sure within those four or five, you know, one of them, you know, could be the breakout one that people come back to and using. And I mean, that's the process we've been doing. Speaker 1: I'll give you an example. I'll give you a quick example. They are either worth zero or a shit ton of money or somewhere in between. Let me explain why. I sat down and had a chat with, I think her name is Anna, works with Fraser Smearton, right? Fraser does 50 million a year in the UK. Big Amazon seller. And it's the geodecoder, right? What people don't understand is that It doesn't matter what you do to optimize your listing, keywords, anything. It's all relevant. If you're mapped wrong and let's say that you're meant to be in baby but then you're in home underscore improvement as your GL, that matters. If you suddenly go off a cliff and 90% of your sales are wiped out and you don't know what it is and you're changing keywords, blaming your PPC agency, blaming your team, changing your images, da da da da, and it's that, that tool is worth a shit ton of money. A shit ton. So this is to say that We take that 50 million phrases and basically they had a problem with their GL but didn't know how to work out what the problem is and how to fix it because they didn't know it exists because it's not something that gets talked about as much, right? But there would have been a time for them, and I don't know the actual numbers, but if you imagine there's a part of their catalogue that they have problems. So every day that that group of products is doing 5, 10 grand a day, I'm just guessing because it's 50 million, right? Say 5, 10 grand a day, that's 5 to 10 grand a day you are losing on a daily basis. But then if you use that free tool and all you've got to do is cut and paste the source code and it will tell you whether it's correct or you'll see whether it's correct or not. That's where that value comes in. So it's worth zero or a shit ton of money but you've got to understand it exists first. Does that make sense? Speaker 2: I mean honestly this is part of the reason and not to plug ourselves this early but why we wanted to move to freemium because you see You see what people are using and you know where to invest your time and that's valuable because this space moves so so fast that actually the key question becomes, What brings value to customers? And you know, what are people using and find useful? Not like, what can I build? Because you can build anything, right? And like, especially if you've got a decent tech team who are motivated, you can build anything really quickly now. So yeah, the focus is actually what's useful and therefore getting stuff into the hands of people, as many people as possible and seeing the use. In my view, it's the best way to, you know, as I said, we wanted to move to freemium so we can move even faster. Speaker 1: So what does freemium actually mean, right? Because you're more for enterprise larger brands, right? Because you're a premium service. Speaker 2: Correct. Speaker 1: So when you say freemium... Speaker 2: Yeah. Speaker 1: Sorry, go on. I won't disturb you. Speaker 2: So there's kind of two things we've been doing in the content, which is giving customers Visibility into how they perform on AI search, Cosmo, Rufus, you know, how they're appearing here. And then there is the actual generating of product listings, you know, optimized for those analytics and the data that we have on the customer and their audience and everything else. We made the decision to give that first bit away for free so customers can log on, they can connect their Amazon account, and they can get analytics on, you know, how optimized the catalog is to Cosmo, you know, their visibility in Rufus, this kind of stuff. And then if they want to generate anything, then they pay. Speaker 1: So it's like an audit then. So is this Cosmo ready? Is this X ready? Okay. Speaker 2: And honestly, I've seen so many I mean, part of the reason that I did it, it's a bit like what we said, right? You know, I made that GPT Cosmo audit a year or two ago. And everyone, I still get LinkedIn messages saying, oh, it's not quite working. It's not quite working. And I always say, yes, it's a GPT. GPTs hallucinate. They don't know all your data. But the customer doesn't want to, they don't know that. They don't care about that. They just care about why it's not working. So I was kind of like, fine, let's give them like the most comprehensive order that you can get, you know, where we will actually, you know, what we've been building for years and give people that and then they can make the decision. And some people, I think, will make the decision that they want to generate content the way they've been doing it before. So they'll have They'll have graphic designers and they may not have that many SKUs. They may be using other software. They may not need to use us and that's kind of fine. That's the gamble we're making is that the customer will trust the data because I've seen a lot. We've got a GPT. I know GPTs don't really work. I've seen lots of other people do lots of other audits for free. I won't name any names. I think some of them are good. Some of them are absolute lead magnets. The context has changed, right? The way that the industry just moves so fast, I think we needed to experiment and get in more people, get them playing, see what matters to them, but also move the payroll slightly down the stack. And fortunately, we've got You know, we've got, you know, money to gamble and experiment and see how it works, right? We can afford to run GPU costs for hundreds of thousands of sellers and, you know, see if they convert into paying customers or not in the long term. Speaker 1: Yeah, yeah. So basically, it's an audit, kind of gives you The checkbox is to where you are with it at the moment and then obviously the next step would be to move on to the pay plan and then do the rest. No, I totally get that and this is what's interesting as well, we can cut this if you want Max, is that you're one of the first in the space to invest in this area and There's a group of us that have been pushing the science into things. Me, five years, being ignored and mocked for about four of them. And then you've got Warner, then you've got Andrew Bell, you've got Joe, Ritu, well, Ritu is a Hall of Famer for the future, and Vanessa Wong. Who else is there? Yeah, there's a few of us. And I think we are in the gap and things are changing. And when you take a little look on Clown World on a weekly basis, you start to see The shift in real time where it's like it's the gift that keeps on giving so they'll go Rufus shit this year Yeah, you should do it and then what happened is more progress more progress more progress Yeah, and so we we all look less like fucking idiots. Yeah understand what I mean, but we're not we don't care about Being right, we care about getting the truth, right? Speaker 2: I know. I mean, I love it. When I had you guys on the podcast and on our podcast, I always used to ask people, like, what percentage of Rufus, you know, will people be using Rufus? Like, people will hesitate to ask, but it's just so clear, the direction of travel. And it always has been. And that's kind of why we got into this because, like, I knew in In 2022, when we launched the business, I knew it was a busy space. I worked in Amazon. Amazon were, back in 2019, looking at Jungle Scout. I was literally in this team, looking at Jungle Scout and going, oh. Speaker 1: For an acquisition? Speaker 2: Not as an acquisition, to copy. Just to straight up copy features. I remember sitting in a meeting, looking at Jungle Scout, and at the time I was working And like the brand analytics, what's now like the brand analytics where you can go in and search query performance, all that stuff, right? And Jungle Scout had done a better job. And of course, like Amazon, you know, is aware of all this stuff. So I was aware of it, but I just felt then that with, you know, the launches in AI that happened in 22 with, you know, large language models and diffusion models, Not only would this mean that new experiences were possible for sellers that just weren't being met, right? Like this is a brand new technology. It's like It's bigger than the internet in many ways. So it's kind of like saying, oh, there's all these established retailers here, but there's now the opportunity to become an internet retailer. And that was my view, like, yes, there's these established kind of players doing stuff for, you know, the traditional search, but there's a way to serve customers regenerative, creating content that wasn't possible. And also I knew that What Amazon would do with the technology would also rapidly change. So not only could like a new software company serve customers and deliver new experiences in different ways, but Amazon was also going to adopt this in some ways. And what they were doing would mean that they'd be, you know, we didn't know about, I didn't know about Cosmo obviously in 2022, I didn't know about Rufus, but I knew that Amazon was going to adopt this stuff in some way and it was going to be massive and it was going to be transformational. And therefore, not only could we build something that would serve people in these new ways, in a generative way and, you know, any content, you know, we were largely generative first, like doing images and text, you know, far before many people. And I think we're the first to ever, you know, to do the image stuff and then text, you know, we were one of the first. But also like how to adapt with Amazon. But what I honestly didn't see is just how quickly things, like from that point, things have moved at like a breakneck speed every single day. I wake up every day, I look at LinkedIn, I look at WhatsApp, and it's like, GenSpark has dropped, this has dropped, DeepSeek has dropped. It's crazy. And also the sellers At least the ones I talked to are really clued in and I honestly didn't think that would happen, that sellers really understand these tools. They've got complex, a lot of the big ones have got complex internal workflows that they will be using. I'll take Sim, because he's talked about it publicly before, on your podcast and on my podcast. He's running an eight-figure brand. They are using the different tools, using GenSpark for research, they're using MidJourney, maybe I think it's ImageGen, probably nowadays, but it's slightly better. For the images like so many sellers have adopted this stuff so it's a it's a like it's a race just to go faster and faster and faster and faster and that's the only way that You know you can you can kind of keep a float in this in this world, right? Speaker 1: That brings me around to another point right is that I have and I in locally via railway versus Versus directly in terms of hosted in the cloud 60 a month or whatever it is. And I put N8n in the same category as That time where you always had a list of things to do, never got around to it. So a perfect example would have been a few years ago, chatbots. Every seller wants to do external traffic in chatbots. It was always on the list. They didn't get around to doing it. And then you've got N8N, right? And N8N to me, and I say this respectfully, this is not reflection of anyone else. An 8N reminds me of the dad up in his loft with his model railway set or his Scalextrics that just goes round. So you're watching all the points, right? And that's not conducive to someone like me who's got ADHD who can... Speaker 2: This is an automation tool, right? This is like a workflow automation tool. Speaker 1: Yeah, so NAN is Workflow Automation. What I'm saying is it just reminded me of, do you remember, we're going, I'm older than you. You know, years and years ago, the kid in the loft with his Scalextrics going around. Unknown Speaker: Yeah, I like Scalextrics. Speaker 2: I know that. Yeah, you press a thing and it goes. Speaker 1: So I'm thinking of it in that context, right, in the offline world. I look at that and I think, I don't want to sit there with my choo-choo train set and I don't want to be that person sitting there reading through a shit ton of documentation, which you should. Because people like developers, going back to the VARC model, visual, auditory, reading, kinesthetic through doing. So I try and work on basis of I'm not a structured person. Like if I write a list, There'll be 50 different things on the list. As it comes out, I'll write it down versus going to click up and having a list for each things that separate that business, what that is over there, design, da, da, da, da. It doesn't work for me. So I need to use NAN and I thought, okay, how can I do this? So basically now, I do it all in Claude. And I have two outputs. I have the JSON file, which then has sticky notes for the connectors that need to be connected. And then I've also got a wireframe that outputs to the artifacts window that looks like NAM. So basically, I can say, I want to build X, this Y, Z, blah, blah, blah. I want to build all these things. And part of the description for this is basically saying, I want you to find the best options, what to use here, you know, the best models or how they should be configured and all of those kind of logics in there that you have gone back and forth with. And now I can write one prompt. He then spits out to the artifacts window, showing you like a mock-up that looks like NAN, right, with a dotted background and everything else. And then he creates the JSON file, drop that into NAN, and then when that loads, it also has sticky notes for what you've got to plug in and where, so it's self-learning as well. So I've had to build stuff for me to use it the way that I want to use it. Going back to my point about AI, like a lot of my projects, bots and everything, everything I build now is keep telling the LLM, you know, build for someone with ADHD, build for someone who's a visual learner, right? And if you look at a lot of programmers, most of them are, you know, when they're using Their model of learning, a lot of that's going to come down to what? Reading and kinesthetic, doing. Okay? Unless they're a UI front-end designer, which is visual. But they're always going to have their heads in manuals or help me files. Do you understand what I mean? And that's not conducive. I want to do this, but I can't do it that way because it's not conducive to where I learn. And that's going back to the conversation earlier is like, how do we get these tools to sellers that you want to do but it's not conducive? One, time management. Two, what output you're going to get? And three, does it sit with your way of learning? Because some people might watch YouTube videos. Some people will want to read and do quizzes. It depends. Do you understand what I mean? It's trying to work out what's best for you to get the best result you can. With the least path of resistance. Speaker 2: Yeah. Speaker 1: So I know I went down the houses down might cut that but the point is is NIN is one of those things where we should all be using. But a lot of people won't because there's the learning curve with it as well. And so they're going to miss out on that unless there are more tools that wrap around it to make it easier for the everyday person with different skill sets. Speaker 2: Yeah. And I think that We're going to see in the next, I would say by the end of the year, people starting to drop, I mean the base model provider starting to drop truly agentic things that you can build. So again, The experience that, you know, software providers can deliver for customers or those who are advanced, like, you know, like sim or, you know, people who are playing with this can build for themselves and in like, you know, in a no code way is going to change. And we're going to have these agentic Models where you can kind of give it a task and it will go and execute that. It will act as if, you know, as if it's a person behind a computer, but there's obviously it's just an AI and it can log on and We do multiple step processes and we already see this a little bit of the deep research stuff. Like we kind of, if you put in, I think every model basically now has deep research, open AI, Gemini, perplexity, all of these guys are deep research where it will take multiple steps in the chain to kind of research and find a problem. And it's really interesting watching it. You know, if you read the, logical steps it's taking, you can see how it's kind of thinking like a human. But I think we soon see that open source, you know, deep seat being the first one, but more, more of that kind of open source, people building on top of that in a very frictionless way. And then that can go and do a whole host of things that that aren't possible today. Like today, yeah, you can generate a product listing, you can generate some images, publish it to Amazon. And that's kind of, you know, what we do. But If you think about all the things that we need to do in the business, a lot of that will be unlocked by agentic AI. You're running an Instagram account or TikTok or whatever, and you want to comment on the specific relevant areas that are in the business you sell. Let's say you're selling dog food and you want to comment on TikToks of dogs in a nice and friendly way that's relevant to each individual post. And then you want to send them a message. And then if they have lots of followers, you want to send them a free sample. Like that, you know, I'm just made that up on the spot, but like a whole workflow will be very easy to do. And I'm very, you know, we're thinking about that. We're monitoring that and we're excited by what, you know, how our business is going to shift because, you know, we've always got to stay two steps ahead. So how we're going to be delivering that to customers in the future. Hmm. Speaker 1: So let's go back on to the point of software, just because I don't think we've finalized it. You've raised money. You're one of the first AI, let's call them AI softwares in the place. Speaker 2: We are an AI first company. Speaker 1: That's what I mean. That's what I meant by it. Not an adaptation as part of the methodology that the technology is built on. What do you think stops someone like yourself in terms of the protection mechanism? Obviously you've got a patent. That's one thing, but the next legal cost and time effort to defend. Where do you see it for you without putting you on the spot? Speaker 2: We have two patents in the business. I think patents will be useful, you know, should we come to sell the business because the buyer is, you know, they can see the tape they're buying, but also If we were to open up our books to a, you know, a larger company and they would not buy us and then in a few months time launch something similar to us, you know, the patent gives us protection, you know, both on the upside of them selling the company but also if they don't buy it, you know, we can go and take action basically for like showing them like all of our workflows, all of our unique IP, everything. So, I mean, I think that's That's why we have the patents. Like, realistically, I know companies have been in, you know... I have done stuff similar to us and it's not like we're going to go, as you say, go and launch legal fees against these companies because it might be slightly different. Who knows if we're going to win? It's a waste of everyone's time and energy and effort and I'm sure it'd be draining. That'd be faster draining my bank account than the freemium stuff which we're doing at the moment. So how do we stay competitive? Honestly, part of it is the culture of the company. We are a very innovation first company. We're 15 developers and three AI researchers and we plan on a two-week horizon. So basically every two weeks, I will sit with my co-founder and go like, okay, what are the three things we're going to build next week? Often this will come from customers. Customers ask features, we'll build them. Sometimes we will take big bets ourselves like freemium. And sometimes, yeah, it will take obviously longer than the two weeks. But broadly, we're reprioritizing a lot of resources every two weeks. That's our advantage is that we move fast. We understand the industry. The second thing is our data. We have lots of data now and we really understand the Amazon algorithms in a way that I think Other people don't and that's because we have good data and we are, you know, working with some of the biggest Amazon, like, globally known brands, like, well, like, Amazon News Us themselves. Zappos is one of our customers, right? So we have lots of data in terms of how the Amazon algorithm actually works that we use as is, you know, in building, like, our second patent is basically a It builds digital twins of algorithms. So we so we kind of understanding how algorithms working in certain categories. So we understand how Rufus will talk about Pet food, whatever, right? That's kind of the technology. And that is fueled by data. So we have a data mode for sure. And it's a specific type of data as well because obviously we've been generating content since 2022. So we also see how AI generated content performs. So we have, you know, performs on Amazon. So we have that unique, unique data of like, if ChatGPT is generating an image for your product listing, it will do so and it will be beautiful, but it will kind of be random. Whereas we know what a good AI generated images for the dog food category should look like to increase sales and conversion. We have that data. So we have a data mode, but really I think it comes down to the culture of just moving quickly, building stuff, and just keeping really short feedback loops because I know that every two weeks something new is going to happen. It's been that way for three years, so it's kind of pointless to If something is going to take us longer than two weeks, we really sit down and think about should we do it. Like freemium obviously takes us longer, but that was a bet that we really cared about. But if it's a random feature that's going to take us longer than two weeks, maybe we don't need it in two weeks because DeepSeek's launched. So, like, why the hell would we... So, yeah, that's kind of... Speaker 1: Ah, okay. So, basically, if I get the rhythm of what you're saying there, He's not only you've got your proprietary technology, you're also plugging into every available. Output, API, whatever you want to call. You're hooking into other systems if you require. Speaker 2: Yeah, of course. We've got two patterns, but we use the base models. We're using OpenAI. We're using ImageGen. We're using Flux. We're using Claude. We have one person in the team whose basically their job is to look at all the AI models, new ones coming every week, and assess should we swap This AI model into our software because it's better. Speaker 1: So basically, it goes back to the point, you know, you've raised seed money, you're independent, you're not backed by, what word am I looking for? Like large investors, what might the word I'm looking for? I have to cut this. Private equity and what's the other one? Speaker 2: Venture capitalists. Speaker 1: That's what I'm talking about. But then what we're looking at now for someone like you, whilst you have to raise money, you're taking risks and stuff. Speaker 2: Well, we didn't need to raise money. So we were, as we talked about, like my co-founder... Speaker 1: Oh, that wasn't, Max, that wasn't a negative. You should raise money because you never know when you need it. It doesn't mean you have to spend it instantly. Speaker 2: It's because, yeah, we have big ambition and we wanted to take big bets, one of them being freemium. Like, you know, we couldn't, it would be impossible to, like, the GPU cost of this stuff is super expensive, as you know, right? And if you think about what it costs, roughly, if you're paying $20 a month each user of GPT Pro, I mean, that would honestly roughly compare to what we would pay for each user that comes on for free, right? It would be like $20 of cost. It's not marginal. Like every new person that comes on, it's going to be a high GPU cost. So no, but we were profitable. For long periods of the company, before we raised and like we're very close to it again, but I mean, it's yeah, it's for the it's for funding the bets, right? It's the funding say, like, let's do something crazy and see what happens, you know, can we. Speaker 1: The point I was going to get to is startups have always been, since they've been around, has been using APIs, third parties, as part of their service anyway. That hasn't changed. The thing is, what has changed is now we are moved in to a new realm with, you know, AI of today, not the same AI that Ellis trained and become a doctor 20, 30 years ago, whatever it was. The point I'm making is that You're almost in the middle of being an alchemist. So basically, you can be a smaller company, take advantage of all the advancements of other companies that are spending billions, and then you alchemize the best parts for your output for your company. Yeah, correct. Speaker 2: Plus our data, plus our patents, and then, as I said, like our culture and domain knowledge of Amazon. We know this place very well, and we have a culture of, you know, like, we're going to move very, very quickly. I felt like we have nothing to lose, like some of the incumbents do, you know, for a long, long time. And we won't say who on the podcast, but I was talking in, doing my European tour, and one of the incumbents, a senior person there said, oh, I haven't changed my keyword strategy since 2017, live on stage. And like, that's not us, right? We know things are moving quickly. We're going very fast. And we don't have like, you know, an old system built for a previous generation of deterministic AI search, which Amazon have moved off of now. We're not encumbered by saying we could, you know, if we were honest to our customers and said, the way you should think about optimizing for Amazon has changed and actually you don't need this tool anymore. You need something else. They're not going to do that, right? Because it's a classic innovators dilemma. It's going to cost them a lot of It's going to cost them their entire business. So we don't have that, you know, and we're happy, you know, we have the view of we'll throw everything away tomorrow, like genuinely. Like, you know, if something changes, we'll throw it away, we'll start again, we'll move very fast, we'll deliver the best experience for the customer. Speaker 1: Yeah. Indeed. So, yeah, going back to the point, will software be eaten up? There are going to be cases for it and it depends if you look at Let's call them the top three, top four, top five, and then there's the rest. You're not in that pack because you're an AI first company. I'm just talking about legacy software. So it really depends on what they have as their IP and their unique selling point and whether they've got certain aspects. If you're someone like Product Pinion, then sure, someone might be able to copy some of the features, but they do not have the market and the velocity and the volume of quality shoppers that output the quality results. So if you're doing Example customer objections. You need a wide range of customers and different territories and you want to get the information fast which means a lot of people because the more people you got inherently it's going to be a bit faster than having a fraction of that size. So whilst on that side that'd be copied, you've got a patent, you've got all of that data that's been collected before this is really taken off because you're at the beginning. So you've got value there. Do you understand? But if you're making software today, And all you're doing is a bit of listing optimization and you're plugging in Claude for an API or that stuff quickly gets replaced. Do you see what I mean? But you need to have a core competence within your business. And again, with the larger companies that, yeah, you might be able to go on to copy my site or whatever it's called and then make YouTube, right? That's great. But unless you understand legacy code, databases, Q&A, storage, setting up environment, all that kind of stuff, that still counts in order to have an application. That's actually going to work and function and then you'll work out how you bring over legacy code and how that works today and it's stable and authentication. So software and engineers stay around because that's the main knowledge. Because don't forget when you give that the same AI platform to me and you, to a world-class engineer, who's going to get the better output and volume? The world-class engineer puts them on steroids, right? We just get a better output, me and you personally, as non-engineers, yeah? And that goes for all domains, I believe. The domain knowledge wins, always. Speaker 2: I do think that... I don't want to disparage new entrants. I think that, as you say a lot, there's space for everyone. And I do believe that... And I think one of the promising gifts, you know, there's going to be many challenges that AI has. On the workforce, you know, think about robotics, think about automating everyone in a warehouse, like that's going to be happening in five years. Every delivery driver, every Uber driver, there's going to be huge changes that happen in the next five to 10 years or 15 years or whatever it would be because of AI. Speaker 1: Yeah. Speaker 2: We will have to grapple with as a society, but one of the positives I really believe is Lots of more entrepreneurs, lowering the barriers to start something, able to build a business that sustains you, you know, to a very comfortable level. And, you know, I'll say this, like we, You know, we before any VC, we could have, you know, not hired the extensive team we hired and the top talent we did and taken bigger salaries than what I was earning in Amazon and kind of like had a nice life. I know who knows how long we had to survive for because it's a fast moving space. But if we pick something which is less busy than Amazon software and gone to something else, I believe and hope that people will be doing that. So I think it's a great thing. It gives a lot more people more purpose in their lives. Speaker 1: Yeah, I think as well, it's not saying it will slow down, but right now it's been There's two things. Normally you get one of these things a lifetime. Mine would have probably been the internet, right? I'm 50 this year, so I'm older than all you guys. But now I've got number two. AI is that next disruption of magnitude. It's different when I talk about the evolution of music. You're going from You know, vinyl, CD, mp3, blah, blah, blah. And that ripped apart the music industry in terms of changing the format. But what didn't change, it was still music. And people still want to listen to it or not listen to it. And people make judgment. They're either like it or dislike it, you know. But I think The opportunity now for anyone, if I take it back to the basis again, you could be a great prompt engineer, you could do all these things, but you may be weak on ideas and asking the right questions. It's a bit like being really good at ranking but have shit products and you're using your really good ranking skills to prop up the shit products. What is it better to be? Someone who's good at making products And then fumbles their way through and then hires people that are good at that stuff. Or is it better to have a shit product and be really good at ranking? I know what I would choose. And it goes back to the same idea with using AI. People listen to this and go, fuck, I'm going to be replaced. You're not going to be replaced. Because if you're an entrepreneur, what do you do? You have a vision. You have tenacity to see it through. And you're very resourceful. There are three key things. Speaker 2: I try to be honest with people. I do think a lot of people are going to be replaced. Honestly, I believe that. Speaker 1: They don't have to be. Speaker 2: It's their action. They don't have to be because some people are going to be 10x by it. I see this in some of the companies we work with for free content. We onboard them and they don't want to use AI-generated A-plus content. They're like, oh, I've got my system in Photoshop and whatever. I mean, the reality is that it's coming, it's going to change the workforce dramatically. And the work of, you know, one person can, you know, it can now be the work of 50 people, whatever. So the reality is in 10-20 years time, you will have less People doing that role, but they'll be doing a lot more. So people have to think like they have to get on board and like Sim has talked about it, the designers in his company who are using this and they're the front lines of this and they understand the latest models and they're coming to him with new stuff. Those are the ones who are going to win and they're going to get They're promoted and they're going to, you know, build a team and you want to build the culture around those people. So it's one of those things where you just have to adapt to it. And like, obviously, as an entrepreneur, it's an amazing opportunity. For entrepreneurs, it's amazing. But even for those who are not entrepreneurs and building companies, but working companies, it's important, you know, it's an inevitability. It's clear where, you know, the genius out of the bottle, right? And you can either be one of those people who embrace it fully and lead the company forward with adoption and trying new things and spotting the next gen spark and becoming very valuable in, as you say, an alchemist, or you can drag your feet and eventually see yourself, you know, be let go. Speaker 1: I had this meeting with my team this morning. There was four of us on the call. And I said, this is your strengths, this is your weaknesses. And I now need to build you stuff to keep up with me. So for design, because I've got an amazing designer, and people see that over the last 10 years, and good designers are out. And I sat him down and said, half these tools I've been building are for you. Because what's going to happen, you're going to be the bottleneck. And I'm going to load you up with a shit ton of stuff. And I can't have the old version of you. I need the new version of you to make this work and scale. And I said that to two other people about their position as well. There's four of us on the call on on that side, not on the agency, but on the events and podcasts and everything else. So I've now got to I've started to a point of velocity with everything that I'm doing and putting everything in place. Now they have to Do what I need them to do at scale, but then give them the tools to do it because all that's going to happen is I was the bottleneck. I've worked out how I am the bottleneck. I took care of that side to a certain degree. But now when I fire off a load of stuff, Stuff that used to take a day, half a day, is going to be done in a fraction of that time. But they have to be on board with that or they're going to slow me down and we'll get nowhere. Because all we're going to have is a bottleneck the other way. Normally the CEO or, you know, The director or whatever is the bottleneck in companies. That's what quite often happens, right? And then they have to find a way out to take it away. And so I've been a bottleneck. And then I then flip it, then my whole team becomes a bottleneck unless they're able to scale everything and do things a lot faster and use their brain instead of, and their domain knowledge, instead of all the laborious stuff. So by the end of this week, as an example, we don't have bookkeeping in terms of the work from invoices and receipts into the inbox, into Xero, into real-time dashboards. So by Friday, 25 hours of labor evaporates. All happens in real time. And I ain't gonna fucking log in, find invoices, forward that on, they download it, label the file, upload, all gone. So we built tools with the engineers as well. So basically, I'll just say what 25 hours a week was that? 100 hours a week in labor. That doesn't mean that they're replaced because there's more volume coming through because I need domain checkers to do that work. So whilst you drop here, it will pick up and their new position with their domain knowledge changing, moving the labor out. Then it's going to come a point where that starts to expand up again. Unknown Speaker: Yeah. Speaker 1: Does that make sense? Because you've got more throughput. Speaker 2: Yeah. Speaker 1: Anything we've not covered? Do you want to finish off in a couple of bits? Speaker 2: No, I think that's been extensive and good. So yeah, I think my message takeaways for people is As always, experiment with AI. As we've said, it makes people more productive. Check out e-content freemium because I'm really proud of it. I think it's very cool and it gives a lot of value away for free. And yeah, give me your feedback if you try them. Just WhatsApp, you know, I'm on the WhatsApps and LinkedIn and whatever. Just let me know. And I'm interested to hear the feedback from people. Speaker 1: Excellent. And I'll drop this in for now. I'm not going to cover it, but it's something that maybe you and I get a couple of the other guys on to talk about is agentic browsers. I watched Sank at the weekend and I've got to set it up. If I'm right and we can discuss this and I don't want to talk about it now until I've got more information together, that's going to be another massive shift. Speaker 2: Yeah. Speaker 1: If it gets adopted, so everything that we're doing now shifts again, because we talk about the customer journey, the customer objection, right? If this is doing all your shopping on Amazon, which one of the videos showed you it did, looking for the best deals, you're optimizing now for the customer. Speaker 2: And the agent. Speaker 1: And the agent, right? Speaker 2: Yeah, I mean, I think, yeah, we could probably do a whole episode on this. Speaker 1: We'll do that another day. Speaker 2: I won't comment. Speaker 1: Yeah, I was just, I put it out there to you because I see it and I'm like, wow, and I've got about 15 million things I have to say about it. But that's for another day. So we'll do another episode in a couple of weeks. All right, guys, thanks for joining us. Sorry about the, there's been a couple of gaps in For the last couple of weeks, I think that's the biggest gap I did in eight years since 2017, but we'll be back more regular now. Now the chaos has calmed down after Seller Sessions, etc. Next up will be a broad match show with me and Adam Heist. Hope you're well. Take care of yourself and your family. Much love and I'll see you again soon.