What's up, DTC Pod? Today we're joined by Frank FERRISSEY, who is the CEO and co founder of X gen AI. So I've been really excited for this episode for a while now. Frank, I know AI is all the rage in just about everything, but we've been chatting for a while and you've been building in the AI space, in the enterprise retail ecommerce space for a while now. One of the most cutting edge products working with some of the biggest brands on the planet. So without any further ado, I'll let you kick it off. Why don't you tell us a little bit about yourself and XGen.

Awesome. Thank you.

And thank you very much for inviting me. Yeah. So XGen's journey started really in earnest, like early 2020 2019 with inception of kind of concept. It's really simple. We're essentially providing enterprise brands with complete autonomous autonomy behind the deployment of machine learning systems and pipelines to their challenges that they face in the ecommerce domain. All these companies, single brands or multi brands, do not have the capabilities to hire their own dev teams to deploy real modern ML tools to their problems. So our platform really enables them to do that without any or hardly any spin up at all from their side and driving immense value. So start off as just like you said, kind of a complicated discovery process and sales process because people didn't really understand what we did. We kind of caught the tailwind here pretty nicely and people are jumping on the bandwagon at a rapid rate. So it's exciting times, isn't it funny how that happens.

It seems like you're trying to sell something. It's not in vogue, it's not what everyone's talking about, and all of a sudden you catch the right wave and everyone's like, wait a minute, you're the AI guy. You've been telling me about this for a couple of years now. Let's talk.

Yeah, 100% tech is super fad driven a lot of time on the consumer side, or B to B consumer side, right. Customers, they hear keywords that trigger them into a buying pattern. And if you're in the right place at the right time with a product that actually drives ROI, even in the down market, you're kind of in a good place. So just going to keep pushing.

Absolutely. So one of the reasons I was really excited to have you on the show is because there's a lot of people these days that are talking about AI. A lot of different influencers creators and all this sort of thing. And we've been waiting to do a proper AI episode in ecommerce, but I said I'm not going to do one until I talk with Frank, because Frank knows more about AI and ecommerce than literally anyone on the planet. So what I'd love to maybe not, but appreciate that what I'd love to do in this episode is kind of use this as a little bit of a crash course in terms of educating people about what AI in ecommerce is, what the capabilities are, right. How ecommerce operators should be thinking about using AIS within their store, and maybe some of the capabilities, what it's really good at, what it's not so good at, and everything like that. And I think you're the right person to do it. So why don't we just start with, at a really high level, what is this step function we've seen with AI and machine learning? If you could just explain the technological developments that have led us to be able to employ AI into specific use cases like we're starting to see.

Yeah, I mean, I think that this has been going on for some time, right. What we're seeing now is, let's face it, it's driven by OpenAI recently, the kind of genesis of this massive global surge of interest. But it's been going on for quite some time. Like, even when I started, certain cloud services were not even up to par to do what we do today. Right? Or if they were, the cost to the ROI model for us internally was not optimal, meaning it wasn't even effective to build a solution back then with certain tools. So I think this has been going on for quite some time. And what we're seeing now, this kind of push is obviously generative AI, right? It's the capabilities to generate text, images, whatever you're doing, with a considerable increase in accuracy towards what a human would do.

Right.

The thing people have to understand is the model or the heart of AI systems. The model is what everyone calls the algorithm. It's what the actual artificial the intelligence is behind AI. There's advancements happening on this all the time.

Right?

MIT is all over research on this. Constantly, constantly producing white papers. OpenAI is a big leader in the space. There's many, many different entities that are doing this. And I think that large language models and generative AI has definitely had a massive lift from recent developments from OpenAI. But what I feel like is this is the first time that you're truly seeing general population accessibility to artificial intelligence in a way that never seen before.

Right.

I think you can look back several years into the past and look at certain systems like Google or sorry, DeepMind cried by Google DeepMind's AlphaGo system. I mean, that was absolutely groundbreaking. The Atari project that preceded that, where reinforcement learning was capable of playing video games on the Atari console. Video games, right. Autonomously. Just give it the inputs, the outputs, and make it learn how to increase the score.

Right?

But that's not something that is like that fad driven. And now all of a sudden, you lease a platform. Anyone can go and chat with it. It's like, oh, wow, Skynet's coming.

Right?

So I think yes, there's definitely not to knock on the advancements because they're obviously huge. Right? And it's really impressive what's occurred recently. But I don't think it's anything new that we've seen such big advancements.

No, absolutely. And that's kind of what I wanted to get into in terms of the language models, right? Like, language models are a huge consumer use case because now you can interact with them. And everyone's starting to see all the consumer use cases that are enabled by these language models. And I think when we were at GPT-3, we saw GPT-3 work. It worked well, but it wasn't like, still, you needed a whole bunch of human cleanup. And it wasn't like you knew you were talking to a robot. And then as you started to get to chat GPT, when they came out in Turbo 3.5, all of a sudden, that started to reach the uncanny level of, wait a minute, this is just as good, if not better in a lot of cases, than a human working with it. And then we saw GPT four come along as well. So I think it's really easy. Interesting to see, like, you're seeing it. You've been working in this space for a really long time, so you've been seeing all the developments, but then the consumers, right? It hits that inflection point where you've given them a place where they can interact with the technology. And the actual language models have improved in terms of their accuracy to the point where people are just like, okay, now I totally get this. This is just crazy. Right? So let's take a step backward. Let's go into ecommerce. Let's talk about all the different applications of AI in ecommerce. If you're a store operator, what are they? You guys at XGen obviously cover a huge component. One of the reasons I love what you guys are building is because you guys literally use AI to drive conversion, which is half or more of the game of ecommerce. But what are all the other applications as well, right? Language models, maybe image generation, all the other things. If you're an operator at an ecommerce store, how are you thinking about AI and what are some of the different tools that you can bring into your tool set?

Yeah, really good questions. I mean, look, first of all, the reason we started off in the conversion side is because it provides direct ROI when you're selling a SaaS product, obviously, to an ecom team, digital team, direct ROI, visible monetary lift is obviously kind of the epicenter where you want to be as far as product goes. But the way I look at AI, as far as, like, use cases, yes. I mean, a lot of people attempt to group this, right? So you can go predictive analytics, ERP, CRM enrichment, CRM predictions. You can get into the personalization side. So product rex, content wrecks, email generation, text generation for emails. I mean, marketing, when should things fire? But this really can get endless. The problem I see is that people try to group this into succinct categories. And yes, that's somewhat the case. But also think about this. If you take code as a general concept, coding on the front end to build websites, apps, whatever you're doing, and then the back end, if you're a developer, it's really a creative tool to do what you want ultimately, right? Or fix some solution. So as opposed to looking at AI as like, plug and play into whatever you want, meaning, like, okay, this AI tool fixes, blah, this AI tool fixes, this other component on ecommerce, right? The way I think about this is it's really a new tool that these teams need to understand how they operate and how they can use them and use them in a way that best suits them.

Right.

The problem is that no one knows how to use it honestly. No one knows how to deploy it, no one knows how to operate it.

Right.

And that's exactly what we do. We automate that process so they can do that where they see fit. And when I talk to ecom teams, what really kind of sticks to them the most is like, think about if you're a developer and you're building stuff, you can really do what you want, right? And you can do things, you can automate things. Code is amazing at automating process, right? Like, everything we do in the computer science domain is there to ultimately automate, build functionality, et cetera, et cetera. When it comes to AI, your ultimate goal here is prediction, first and foremost, but to building the right systems to automate. So what does that really mean? Well, it really means you're building workers to do your job, right? I say that tentatively because I know it's like a bit of a sour subject these days. But think about this, right? If you're an ecommerce manager and you have X duties on your job description, most of these people are getting like 10% of that done, right? It's just madness. They're overworked. There's too much to do. The amount of revenue going through their platforms is absurd versus how many employees they have against retail, for instance. So to them, it's really about, hey, look, you can finally deploy, train, teach, and deploy workers to do what you need to do either to automate your features and your workflow or to make it better, or to do it at a more personal level to the end consumer than you've ever done before. So use cases surrounding this. Again, there's like hundreds, in my opinion.

Right?

But again, what I'm trying to on a mission to do is really teach the market that it's not a modular approach. It's not like this SaaS approach. It's like, I want a widget that does X and I want another widget that does Y. No. What's your challenge? Cool. Here's the tool to deploy AI systems to your specific challenge. And that's where I think the future is.

And I totally agree with that because I think as we see all these tools popping up, they all might do one really specific thing and one narrow thing, but then you end up with a totally fragmented workflow where you're using one AI tool to do this, one AI tool to solve another problem, and the next thing you know, your work processes are totally broken. And then those tools aren't communicating with each other. And then you're even more confused than you were when you started out. So why don't you talk to me a little bit about maybe your guys'approach at XGen, right? Like, when you're saying you're building the AI or machine learning platform for modern ecommerce teams. What does that mean? What does that mean for an ecommerce manager? How do they use X gen? How do they deploy it in the simplest way possible? What do you guys do for ecommerce teams?

Yeah, it makes sense. I mean, in order to understand what we do, just quickly take a snapshot of the status quo here, right? These teams are used to buying a SaaS solution to solve a single problem, right? It's like our car checkout or abandonment emails need to be automated. Calls buy it all to do that. Our pop ups suck. So let's get something better. We'll buy a plugin to do that.

Right.

With action AI. What I found is obviously any SaaS platform these days, at least 99% is going to include some format of AI.

Right?

But the big difference here is who's controlling these systems, right? If we as a society are starting to put artificial intelligence at the forefront of controlling what we do, what our consumers do, what we show consumers, how we operate our business. The thought starts to cross your mind is, should we really not be seeing what's going on? Should we just be openly trusting another platform to do black box what we should be making decisions on, right? And that's what a lot of the hang up in the past I've seen in these solutions is like. There's no transparency about what's going on. But also when you look behind the scenes and you start to understand the deployment of ML machine learning systems to problems, you also understand that the way you dial these systems in is super complex in some scenarios, right? A lot of people don't understand that AI is at its more advanced stages. It's not like some automated system you just drop in and yay, everything's happy. No, there's a lot of people working on those things to make them happy. Some on a daily basis, some on an hourly basis, right? We're a long way from dropping a robot off somewhere and being like, cool, it's happy, let it learn for the rest of its life, right? In that light, if you're building solutions in the past and you know that complex machine learning can take a lot of a larger workforce with a higher expertise level to operate it naturally in this PNL driven society, you're going to cut corners, right? You're going to look for the option that has the lowest possible cloud, compute component, et cetera, et cetera. So with us, the real power or value that I saw was basically saying, hey, what if? You could actually just expose that to the customer, right? What if you could just show them how to operate it in an incredibly dumbed down way, right? And start by saying, okay, you don't know anything about ML. You can deploy ML systems through our platform. And if you don't know anything, okay, you can basically set it like a black box, right? So everything's automated. But the real value here is as time goes on, especially enterprise trends are typically driven at the enterprise and trickle on down through the SMB in my opinion. Especially in tech, right? So you look at this and you say, okay, so one of the trends I'm seeing is that retail brands are starting to hire AI teams, right? Which is new, okay, and not necessarily big brands, brands doing anywhere from 100 to 200 online. And you start to see two or three data scientists, ML engineers, whatever they're calling them, right? And you ask them, hey, what are you guys doing here? And they're like, well we're just analyzing data. And they're like, okay, here's a tool that's fully autonomous. But if you start to uncheck automation, you can tweak anything you want. And they're like, wait a second, this is what all those awesome developers are doing at the advanced level. But with action AI, you can match your skill sets to what's available and configure it. So that either it's fully autonomous or like 20% or 50% and you're good to go. And the end goal with this is that we push everyone towards full configurability, meaning we want democratization machine learning. We want them deploying systems to their environment where they are fully in control. I do see a world where business in the future where businesses are driven by humans. I believe that is the case for a long, long time. It's just too complex a decision process right now. But for all the workhorse stuff, AI should be being used, controlled, and fully configured by the people who are running those businesses, not by some unknown SaaS company that's really just trying to fit everyone into one box.

Right? Yeah.

And I think that's also a really good point about what the role of the human with the AI is. The AI is a great way to think about it is the workhorse. But if you're not setting that workhorse in the right direction and you're not tasking it with the right things, like you were saying, the decision processor, there's so many different levels that you go through. And I think you see a good example is like the AI agent stuff that's come out, right?

Sure.

You have all these autonomous AI agents that you can set after a task, and they'll evaluate how they're doing and they'll keep executing, but at a certain point, you almost realize you're like, oh, wow, my OpenAI bill is going way up, and I don't know if I can really use any of these outputs. Right. So I think one really interesting thing that you just mentioned about how you guys are thinking about AI is giving not only the teams the tools to be able to use them, but also you said something very interesting about analysts.

Right?

Data analysts. It seems like every brand wants to analyze data, but taking action on that data a lot of times requires coordination and execution and a whole bunch of other things. So one thing that I think is really cool about XGen, and I'd love for you to talk about this next, is when you guys say that you're kind of bridging that gap between not only understanding what's going on in the data, but also using the AI and the ML to drive conversion and drive action. What does that really mean? Because when I've seen you guys run things, it basically means you can make the decisions and you can help tune in a couple of different things about the platform. But ultimately, XGen is doing all the work for you in terms of generating the sales and converting revenue. So, yeah, I would love for you to just talk about what that looks like.

Yeah, I mean, look, so we launched our tool really in the initial stages, focus entirely on the user experience.

Right.

So what are your shoppers doing when they're coming to your digital environment. And if you look at that world of optimizing for conversion rate, personalization adapting or geolocation, whatever you're trying to do, marketing funnel, post site touch point, not upstream of X, we don't do that. You've really got this like you start to look and it's like there's a huge amount of decisions to make, right? And if you really dig into ecommerce, what's amazing about this is the amount of data you have is unbelievable. Like it's an incredibly data rich environment. So for us, it became like, well, okay, so you're trying to deploy product recommendations, right? And you're coming into an environment where some people are like, oh yeah, we just deployed best sellers, which is basically saying, what's selling the most that day and showing that to all of your million shoppers on your site that day, right. Then some more sophisticated systems came out. But again, even with AI coming into that specific use case, it really became like a clustering of behavior or a regression. Actually, we'll skip that. More like a simplified way of decisioning where someone should sit as far as the experience you show them, right? And this is where it gets really interesting. If you look at what Amazon.com is doing, who has access to basically unlimited resources in terms of their developers and tech, right? They're deploying really deep, sophisticated AI systems, right? And if you compare that to the SaaS category now, it's like you're basically going to tap into a black box to run your product recommendations. You don't know what's going on. And this platform was built six years ago and it's trade secret as to what they're using, what's going on behind the curtain. So the one thing I can tell you is it's probably locked in time. Like it's not much happening in terms of changing over time to that SaaS tech, right? Maybe yes, maybe no. I'm going to vote no. But then you look@amazon.com, it's like, well, no one has the budget to spin up a dev team, but the difference in technologies in three years is huge, right? Like massive. I mean, you should see some of the stuff that comes out from MIT on even recommendations sometimes. It's profound. So the difference here is the time to usability. MIT releases a white paper. You're not going to see that hit SaaS for a long, long time if it is affected, right. Then you start to look at again, going back to Amazon.com, they're like predicting use. Like you got a million users on the site, they're predicting every single one of them uniquely what they're about to do over the next two to four minutes on the site. And they can literally play chess with that customer to get them to convert. Hey, if we show them this product here and there, and then there, and then do it a certain sequence, the probability for that user to buy goes way up right. And that's why they announced, like, hey, 20% to 30% of our rec strategy, sorry, 20% to 30% of their revenue increase is attributed to their rec strategy.

Right?

It's like, wow, so going back to that, existing companies or customers that we talked to do not have the capabilities to do that. Right, and you even find some heroic internal efforts of some lone developer spinning up some actual real time prediction system to do this and have done a great job. But the difference here is, as opposed to focusing on the algorithm, which AI always used to be about, when you talk to these guys, which is the model, it's really about like, hey, you should have a toolkit to do this, right? And those tools should include stuff that came out like a month ago out of R and D, because if you don't have that, you're going to get left in the dust really quick. So our philosophy is, hey, let's stay on top of tech. We know what we're doing. We know about AI technology. We will make that available to you in a toolkit of things that you can use all kinds of different stuff and you can play around and deploy what's best for you. Of course our support team will help you and guide you. But that's the general philosophy you can test this time and time again. Take a modern neural network, right, like a really sophisticated neural network. Plug in all the user data, CDP, customer data, platform data, and then take the entire product catalog. You can literally do things like rank 500,000 products ten times a second to every single visitor on your site, right. And that basically means that 500,000 combinations of stuff is completely unique for every visitor every 10th of a second. Right. We're talking next level stuff at this point, right. So to empower brands to leverage, that is really exciting to me because again, it's out of reach, but with tools like this, and again, Blaine, it's following a trend, right? Like, go back ten years, 15 years, not everyone could deploy ML as a developer.

Right?

Now you've got all these tools where it's like, oh, you got Amazon, AWS, Sage Maker. It's super easy. AWS personalized, like plug and play, back end services you just throw in there, oh, cool, we're getting blah, right? So our aim is to bring that to people who don't understand a line of code.

Yeah, and I think that's a really big distinction because while a tool like Sage Maker, for example, might be great if you've got that whole dev team that you're supporting who knows how to build out your stuff, great, go ahead and build it. But not only is it going to be expensive for you to create, the labor is going to be expensive. Building it's going to be hard. Maintaining it's going to be hard. Keeping up with industry standards is going to be hard. So it seems like at least from what you guys are building, you always hear about the build versus buy argument. This one's like for at least in my mind, this is like the biggest no brainer. Because again, like you're saying, it's not just about that. What your job is, is being not only building all the models, understanding how this game of I like how you put it. You guys are almost enabling brands to play chess with their customers, with the goal being conversion. Right? I think that's a great way to think about it, but it just seems like this is one of those things where it's hard enough for an Ecom team to manage a whole Dev operation, let alone spin up their own personalized recommendation system and then their own AI tools to do exactly what you guys are able to provide. So the next question I'd have on the back of that, I'd love to talk about how this actually works in practice for some of the customers that are using you guys. I think it's really cool because I've had a chance to play around with it and also talk with brands who have implemented you guys. I think one of the coolest things and I'll just break it down for the customers and then you can provide some color as well. But basically the way it works and the way I've experienced it is you will land on an ecommerce site, whether it's the site or a specific PDP. And what XGen is doing is basically based in real time, based on where you're clicking, how you're interacting with the site, it's serving you recommendations in real time. So when I go experience the site of, for example, Valentino, one of the brands that you guys work with, and if you were to start browsing that site, we're going to have totally different experiences in real time based on whatever AI thinks the best way it is to get Blaine to convert and to get Frank to convert. Is that an accurate way of describing the user experience at least?

Yeah. So from our user prediction pipelines, which that sits under, which is predicting shoppers for various purposes in the context when you apply that to recommendations or product recommendations outside of other stuff we do like image recognition or some of the chat tools we use for the recommendation system, basically. And again, it comes down to the model you're using.

Right.

But the framework that we put these models into allows for to solve. Basically the model's job is to solve the problem of you've got a million users on your site, right, and you've got maybe 30 locations across the home page, the category page, product detail page, cart fly out checkout or actual cart page, maybe some personalized experience pages, a whole pile of stuff.

Right.

So in product discovery, the context has always been quite cloudy surrounding like, okay, a lot of the focus a few years ago. Was like, okay, if you are a first time visitor, you should get this experience, right? And that does make sense. But the question is, well, why should I be grouped in the same category as 500,000 people right now, right? And then in that specific personalization domain, we took this huge dive into capturing Identifiers. So let's authenticate grab your email address. Let's actually put a cookie on there. Let's grab some third party cookies across device or across platforms and identify you. And the thing I always come back to is a retail analogy, right? It's like you walk into an apparel store and can you imagine if someone walked up to you and be like, what's your IP address? Give me a credit card before you even look at any products, right? You'd be like, I'm out. The context here is like, okay, let's compare this to real world. What happens in a luxury boutique. Let's call it like Gucci, right? You go in there and the trained sales associate is going to be so good that they're watching everything you're doing. The color of the item you're picking up, the category, the metal, all these things. And then they can educatedly guide you towards helping you to find the right thing, right? As opposed to saying, are you a first time buyer?

Good.

Go through door one, which is like a warehouse full of shit, right? So that context is a little bit not really the way the world should work. So with our models, what you're capable of doing is basically taking all the products you have on site, taking all the interactions from that user in real time, enriching the product side. So, like, what is the color, the category, the weight? All these different things that could lead up to a buyer or a shopper making decisions and predict every single outcome across every single one of these categories to basically rank the entire offering to you from one to 500,000 in terms of products as to what is the most likely to convert you right now.

Right.

And then show that and then adapt that in real time. So every interaction a user creates, it morphs and changes as you go throughout the site. First of all, it's an incredibly powerful tool, right? Because obviously no experience. You have a million people on site. No one's going to see the same thing ever. And also you're not going to the same thing. Every page you go to is going to adapt to morph as you go. But more importantly, in running complex AI models on problems like this, what you start to see is shopping is actually like a chess game, literally, right? It's a series of moves that if you do it the right way, you will bring that person a lot closer to buying versus if you do it the wrong way, you're going to bounce them, right? So that is what these systems are capable of doing is predicting the journey and adapting it for every single visitor uniquely, right in that pipeline or that use case. That's what our models and ML automation can do, right?

No, I've seen this work at scale and what it can do is absolutely crazy. But I'd be curious because a lot of our listeners, right, they're building, they might be starting up a brand, maybe they're going 10 million to 50,000,050 to 100. So maybe on the smaller scale of ultimately scaling these big commerce operations. But I'd love to know, why don't you talk to me about from that perspective? Do you guys support shopify? Have you seen any success for these sort of catalogs? I know you guys are really focused on the enterprise, but just talk to me about a if you offer any solutions for shopify or shopify plus brands, or b what the AI strategy is for those who maybe don't have 1000 products in their catalog but maybe do have a whole bunch where AI and personalization becomes really relevant to the experience.

Yeah, I mean, that's a really good point. Yes, we definitely work with Shopify customers on the Shopify platform. Shopify plus mainly. The thing I would say, obviously we do focus on enterprise and then we do intend to bring out tools for Blaine. The smaller the data from these customers, the less you have at your disposal to use.

Right.

However, that said, what I do find later on for brands perspective who are growing an ecommerce business, what I would implore them to think about is a lot of the solutions we provide, like image recognition tools to fix the product catalog attribution color associations, categories and attributes, analytics and predictive analytics through our services. It's all fixing data issues that if they'd done it right in the first place, they would not have. Right? Dude, I've seen big teams like 30 plus just dedicated to fixing problems that if they've done it right in the first place, there'd be no issue. So while action AI does not service the lower Shopify markets right now, what I would say is that you really want to start thinking about a data driven organization, right? Because if you have that in place, at least when you start to leverage AI components, you're going to be ten times more agile than the next competitor, right? So think about all the data you can capture about your customers. Not PII. Just think more about events. What are they doing? What are they interacting with on a real time basis? Your product catalog, you've laid things out. You start off with 100 products, you grow up to 1000. You're selling golf tools. You've got gloves, caps and golf clubs, nine irons drivers, whatever it is, right? But think a step deeper. What is that golf club made out of? Is it steel? Is it aluminium?

Okay.

Is it titanium? The grip, is that leather? What is it? Right. Start to think that deep, because I'm telling you, when you get into the later stages, if you have that schema configured now, you are going to be ten times better off and ten times more capable when it comes to machine learning than anyone else.

So that one's super interesting. So really thinking about the approach because it seems like a lot of the brands that start up, maybe they start out with one hero product and then one Hero product grows into two, grows into ten, and the next thing you know, when you're a real behemoth brand, you've got thousands of products that you're offering. But it's really interesting to think about all the different attributes, about products that you want to be tracking and setting up your data infra to scale, because I think that's not something a lot of at least my experience. In talking with a bunch of ecommerce brands. A lot of them first scale, and then they're like, okay, when we scale, then we'll start to figure out data. But really taking approach from the early innings and what's a way to do that beyond just what you had just mentioned about thinking about different attributes and ways to set up a schema. Do you have any other insights in terms of if you were to launch an ecommerce brand, how you would be thinking about data?

Yes, it's funny you say this because we have thought about releasing an analytics version of our tool that comes with basic models that work with much smaller data sets, but it really sets the brand up to get their data schema configured so that they're good for the future.

Right.

And primarily Functioning is an analytics tool. It's like super enriched out of the get go.

Right.

But if they just really focused on what the consumer is doing in the environment, forget all the other noise, forget third party signals, and if they're on Facebook, just stop thinking about that.

Right.

First of all, a lot of that data is not accurate anymore. If anyone hasn't noticed anything coming from cookies these days, especially third party are like, you cannot trust that stuff much anymore.

Right.

So I would say really focus on what the consumer is doing. Just think like a retail store in person and you're a shopping person, you're a sales associate. What would you be looking at to sell that person? That's the data you need to be captured.

Got it. So a lot of event data as well from your site, right? Like you want to be really capturing all sorts of events, maybe not just the clicks, but also matching those events with the product attributes and everything like that, right?

Big time. Yeah. Right now they don't need to worry about matching things as long as they have accurate records. Like things are like the time and dates on them are accurate, you should be okay. But again, you got to do both sides, right? So ideally, if you're capturing product data, you do want to enrich the product categories, attributes, colors, all the associated points to do with your product. As deep as you can take that, the better. One of our other AI tools is product catalog enrichment.

Right.

I mean, you can use a really sophisticated computer vision tool that was just recently enabled a part of it through OpenAI's tech that allows you to basically pick your schema. So you can type in any schema you want, come back 24 hours later, and whether your product sets ten or 10 million, it will associate all those components without telling. It a thing you don't need to annotate. You just say, like, we've seen people type in crazy stuff like Golden Lion, and it goes and finds the association of golden to lion finds that in the product catalog. Oh, this t shirt has a Golden Lion on it.

Cool.

Let's throw that into that category. And then next day you can look up analytics. You can say, Golden Lion logo products, sales in Japan last week, it's all there. So stuff like that, right, that's really.

Cool, especially as a brand scales. And the more of these attributes that maybe you weren't even being able to track before. Like you were saying, there's certain things that you would track, but there's other attributes or maybe commonalities between products that you wouldn't even know. And hey, now you've got a category and that converts, so let's do more of that. The next thing I kind of want to talk about is where do we go from here?

Right.

In AI. I think clearly, if we've seen anything in 2023, it's the fact that AI is a very real reality in every profession now and adopting those tools. But for purposes of the pod and what you do and what I do, let's just keep it in ecom for now. Where are we now? If you're an ecom team, what should you be? If you're not doing something right now with AI, what is it and what do you need to be doing? Tell me about who's completely behind the ball if they're not implementing AI and what the bare minimum they should be doing is. And then let's talk about what some exciting developments in the future of AI are going to bring.

Yeah, I think in the ecom domain, I think first and foremost, brands should be really looking for partnerships where they can learn about machine learning. Like, it needs to be part of their company culture.

Right.

You just have to think about it from this regard. If you are a brand and you're building a product and you have a competitor that's very similar to you as far as market penetration, and let's say you're selling clothing and you're like, I hate my competitor, what would you do if you found out that your competitor is converting like 30% higher than you?

Right.

Or that their loyalty program is literally adaptive in real time to every single customer and it's bringing back, it's increasing the LTV of your sales by massive numbers. Or the fact that you can literally dial in your product analytics down to the metal type you've got on your product. And then that's going back to the design team and you're then predicting what type of metal is going to sell in a country. And then design team is building products to suit market demand. Your time is limited, dude, you don't understand the amount of power these brands have that are going to leverage tools like that. They're just going to take over. So first and foremost is start building a culture surrounding ML. Look for services that give you the complete configurability to deploy ML where you want it right under your control. And when I say this, I mean don't be afraid to say, hey, what model are you using behind the curtain? If they say, what's trade secret, I'll just walk away. I'll be like, no, I need to know what model you're using because I want to learn. I want to have a library of stuff that I can pick from and make it the right one for me.

Right?

So where I see this going, obviously my opinion is that first of all, ecommerce is a technical subject, meaning the tech stack that ecom sits on. It's kind of controversial to say this, but it is an absolute insert bad word show, right? It's terrible. It's literally a disaster of technology stacks sitting on top of each other, right? So what I would say is start to think about how you can make your applications or your environments native. Starting with the data. What data do you need to drive the right AI systems? How can you build applications and web pages that are natively capturing the data you need? And this is getting into more advanced development stuff. And then from the ecommerce side where you should be using AI to automate a job that you know that an AI system could be doing better than you, right? Just call a spade of spade. You cannot talk to a million customers a day. AI systems can. So where could you be leveraging it and start working towards implementing those systems? And the other component to this is the data side. Privacy is like tip of the iceberg. I mean, did you see the fine Meta got hit with from European?

Yeah, I just saw that.

Yeah.

I mean, this is unprecedented, right? Like this is next level and the TLDR behind this is guys, data does not belong to you ecommerce people out there listening to this. It is not your data. You may think it is, but it's not. So the concept here is stop thinking about how much data you can capture. Meaning all the stuff across third party applications and all this stuff and just think about the right data you need to capture and capture it anonymously. That's what I would say, like, to future proof yourself. You need anonymized data and you should only be thinking with your ecosystem. Stop thinking about what you can take from Facebook and Meta as far as data is concerned. Just think about what you own as a platform and the data you can grab to make things function. And again, another plug for us, just because I have to say this, because it's super important to me, is all of our systems and AI tools were designed to operate solely off of the customers data, right? Every time you spin up an account for action AI, it just spins up the complete resources like a dev team would for you as a brand with your closed ecosystem, closed database, everything.

Right.

So I really think brands should start thinking about that as like, the central launch point to prepare for the AI workforce coming into the white collar environment.

Yeah, absolutely. And I think kind of what you were talking about, about data privacy. I know a huge thing you guys do is your X gen Works cookie lists, right? XGen doesn't need to know who you are, doesn't have to have cookies for what other sites you visited to make its determinations. It's just like the shoppers showing up to the chess game and X gen's showing up to the chess game, and it's like, okay, let's go. So I think that's really cool, and you guys have done a really good job there. Just as we kind of wrap up here, one last question that I'd have that I'd be super curious about is what do you look to for resources? Like, if you're trying to learn more about AI, right. You've got all these people on Twitter talking about all this crap pretty much popping up every day. There's a lot of noise in the space.

Right.

So if you're trying to educate yourself, learn and have good resources, are there any resources that you found really useful for yourself or that you look to tap into?

Yeah, I'm a little bit different because I can study things that I think most people would not understand. Right. So it's a little bit what I do, you know what I mean? I don't know how to design clothes, but I can do that.

Right.

What I would say is, honestly, Blaine, I think there's a huge gap in white space in the market for people learning AI tools from the most. I even thought about doing an educational series on social just to help people out, right. Because the one thing I'll say is, AI may seem incomprehensible and complicated, but it's really not. First of all, it's not half as intelligent as people think it is, which is really cracks me up. But second of all, I think that most ML experts are like, ten steps too far above the heads of the common of the average person who's not in that domain.

Right.

So I would think that some basic educational process. You can obviously, as demeaning as it sounds, will start googling things like, what is machine learning for dummies?

Right?

And you do get better explanations. I did that at one point in my life. You know what I mean? You have to go through that. But again, yeah, huge white space for education for the everyday person than ML, for sure. I completely am behind that bandwagon all day.

Well, yeah, I'd be down. Maybe we can start creating some really good educational content about it. Because I know it's a lot of guys that were into crypto and now they're AI influencers and it's like, okay, slow down. But I'm really glad we got to have you on, because you actually know a thing or two about this stuff. And for anyone who's listening, where can they find out more about you? XGen? And yeah, are you on Twitter, LinkedIn? Where can they find out about the company? Everything?

Yeah. Twitter, LinkedIn, Frankfarrissey, instagram companies. XGen, AI. Pretty simple.

Sweet.

Well, thanks for coming on. We'll have to have you back and we can break down more in depth AI education and crash course for ecommerce brands.

Love. But.

Thank you, Blaine. Thanks for having me.

Thanks for tuning in and we hope you enjoyed this episode of Dtcpod. If you enjoyed the show, we'd love your support. A rating and review would go a long way as we continue to host the best builders in DTC and beyond. Follow and subscribe to the show and make sure to check out our show notes. Where you can find our socials and weekly newsletter, visit us on Dtcpod.com.