Fabric’s Mike Micucci on Transforming Retail with LLMs

ai video series. I'm your host Mike Ra. Today we're with Mike Kuchi, who's CEO of fabric, and we're talking about how it, and the use case for retail in particular is gonna be transformed by not just ai, but this whole notion of composable services and headless services.

0 experience. That's for sure. Mike, welcome to the show.

Thank you, Michael. Pleasure to be here. It feels like retail has come a long way.

I kind of remember when retail was always considered a technology laggard in, in the last, you know, few years or so. They really, uh, are maybe at the, at the front edge now of technology adoption, and I'm sure AI is a place where we'll see a lot of it next. But what is your assessment of what's going on with retail and technology these days?

Everybody seems more dependent than ever on it and has the whole approach to the way they think about it changing. I they, taking a step back, you think truly through the lens of e-commerce, which is, you know, our specialty here, e-commerce all through, probably pre covid, was really focused on, you know, building out that digital channel. And, you know, COVID just dumped a massive amount of investment into the whole digital space, uh, and accelerated a probably the entire industry, five to seven years, like, just pulled it forward.

And coming out of that, the, the structural changes in the market where your, your customers now are always starting digital first and how you address that customer from, you know, how they find your products, how they make sure that you have those products and ultimately how that you, uh, deliver them has completely kind of reshaped the entire market. And the investments that you'll continue to see is, you know, we always talk a lot about, well there's the, the kind of personalization aspect of how I personalize a shopping experience, but the real, um, real innovations are happening behind the scene and the efficiency of getting products and inventory from point A to point B. Um, and that's where the big investments have.

So yeah, I'd, I'd say, yeah, it's been completely, it's going through the next transformational phase, uh, um, as we speak, coming through what we learned in Covid. Yeah, it seems to me when I look at a website and, you know, maybe I look at that more technically than most, but it feels like they're now essentially a set of composable backend services that people are stitching together. Some are things they provide, some are provided by third parties, and then they're trying to dynamically personalize those services for folks on the front end.

And it just seems like the whole experience has become a lot more dynamic and, and therefore challenging to manage. That's a good point. I I would also maybe point out that we might be, at a term we've been kind of calling peak website because your customers are probably showing up in many different places now.

Uh, they might start their journey on social and maybe search or marketplace, right? So I would say if I broke down the experience, it's particularly if you're selling online, it's around the product detail page and that detail page highly personalized to you. And it's only gonna get more so with AI on how that product is represented to the way that you, the consumer, uh, uh, gets the information you need to buy.

That entire cycle is now really, really gonna be focused on that PDP page. So that PDP page, uh, product detail page, it may show up on your social feed, it may be in Google cer may be in a marketplace, but that's where you're gonna see the big personalization aspects happening. Um, and that's where it's all coming together.

And a lot of it is componentized around a set of very flexible backend services that can be combined in different ways to meet what the consumer wants to do for their shopping experience. I, So where will AI find its way into that workflow? Because, uh, we see these generative AI models, there are large language models, um, but I am personalizing them is a challenge, right?

Because I have to collect data and then expose that to the LLM or extend it in some way. Or might retailers build their own LLMs in the future? How, how is this gonna kind of play out in your mind?

Well, um, first I think we've been working on AI recommendations and pieces like that for years, years, years. What large language models are doing is they're unlocking a set of capabilities that I think are just gonna completely transform the shopping experience. But I'll, I'll get from the, the macro and then kind of maybe more in the tactical.

Um, I don't necessarily think you'll see retailers generating their own large language models, but you'll have very commerce specific models sitting on top of those. Some of those will be on automation of tasks that were very hard to do. Some of 'em are maybe on personalization and some of 'em are just be helping them guide the overall business.

Now, I'll give you the simplest example. I think one that's fairly well discussed industries just around product descriptions, right? So let's say you're managing a catalog of a couple hundred thousand products that you're selling.

Maintaining the product descriptions is difficult. It just takes, yesterday it was manual. AI can automate that with generative ai, right?

You can build models on top of the large language models that tune it to your business. But let's say I wanna market this shirt to two or three different kind of demographics. Well the cost of generating a product description with attributes that targets different demographics for the same shirt, now it's dropped to virtually zero.

So I can have a description tailored to you. I can have a description tailored to a different demographic, a different region, different language, all centered in one kind of core data, even on the product catalog that was just non-accessible thing. And then you can take feedback from the search engines for your site, for your marketing campaigns and further tune that for optimization on conversion based off of the demographic.

And that is like, that is right, that's a big change. But it really just came down to we can now manipulate and change that product description virtually at will. Um, and then save it and use it for all different types of things.

Such one, I think that's like the most accessible early use case that many people see. But you can take that many, many steps farther, whether it's images, attribution, um, and then you can use that across your entire business. Yeah.

Now you folks are working with, uh, Amazon Web services on some of these AI initiatives. So, um, where does that intersect with, uh, your platform in terms of they have these services that they provide and how does that all come together? So Michael, the question around AWS, we've been a long time partner with AWS uh, we're really excited about working with the, the bedrock technology.

Um, we use different models for different things. We use large language models as the foundation models for everything I just talked about from how we, um, model product information, how we automate supplier onboarding so that you can bring more and more products in to really heavily on order management fulfillment, uh, inventory allocation, uh, that allows us, bedrock allows us to select which different large language models we wanna use and also the technology on top of it of how we build our own models that are very specific to, um, the particular automations and AI use cases that we want. This technology gives us that flexibility, but also allows to tap into the scale and breadth that AWS adds gives us a lot of time to market advantages and cost to serve.

Um, it's really flexible and um, you know, we think to just give us a good advantage in the market as we jumpstart, um, beyond kind of the baseline AI capabilities, but moving more into how AI is helping guide our customers to outcomes versus just tactical use cases. Mm-Hmm. So coming full circle to the previous question, how much do you think retailers are just gonna rely on AI capabilities provided by a platform such as yours versus something that they go, you know, contract with a data science team and a bunch of data engineers to build themselves?

Well, I think there's a couple pieces. I think this is a really good question. You get a lot from the underlying foundation models, but to really tune it to an industry, you need to build specific models on top of that, right?

That are very specific to the commerce use case that you're providing. You need to develop the reinforced learning and the other learning techniques that go into it so that you are constantly tuning the main model that's gonna drive your business. So you get good results from a large language model.

But to get optimal performance, you're gonna have very specific models that are built on top of those. And those specific models are trained for your business, they've been learning from your business and they can accelerate overall, um, any particular use case you have. So I believe that a lot of companies will use large language models for generic use cases.

When you get down to very specific pieces, um, dedicated models for their business is, is really what's gonna change the game. Mm-Hmm. So in my mind, not just sellers, but buyers, whether they're consumers or B2B businesses are gonna have their own AI agents that are optimized for buying things.

Um, and somewhere along the line, those AI agents are going to come into contact with AI agents from retailers that are optimized for selling things. So will these two kinds of different AI agents eventually meet somewhere and battle it out until there's some outcome that we can live with? Or will they just eventually send a message back to us and saying, Hey, can you help sort this out 'cause we're at a standstill?

Well, I think we all would like an AI agent to help us shop and get the best deals. Um, and clearly, uh, I think that's gonna be the future. You're seeing how AI is being baked into the devices that we use every day, and that's gonna be an accessible thing from the shopping side of the world.

But from the brand's perspective, from retailers, um, you know, what you're gonna see is much better market dynamics as far as product allocation and personalization that will help meet in the middle. To really do that though, you need a very flexible underneath commerce platform, right? Because you're gonna start doing combinations of things at speeds at which you never had to do before, right?

You used to, the, the kind of previous generation of websites as we talked about, were optimized for people coming in, they shot the sale, they buy what they need. The next generation is commentorial. You might be displaying your product details all over the place against agent A, agent B, agent C, and the amount of, uh, underlying requests is gonna quadruple, right?

So you need to have the independent scale and kind of ability to personalize on the fly that a combination of AI and a highly performant platform built around a cloud native architecture and things like AWS are gonna provide. So this is gonna kind of usher in a whole new level of technology requirements. And that's happening today that, um, I think the previous generation of tech just wasn't built for.

It's a very exciting time. So I don't think the agents are just gonna battle it out, but what's gonna happen is gonna allow shoppers to have a modern consumers have a much better experience. At the same time it's gonna allow brands to represent their products where the consumers are and help them understand their offerings.

I think it's a very exciting time. So what's your best advice to retailers about how to get ready to harness all this? 'cause I think, um, everybody's excited about the possibilities, but I think there's work that needs to be done to make that possibility a reality.

Well, first of all, I i look, we need to focus on basics, right? This is about being efficient. Um, you know, right now consumers are shopping digital and we know all about like good ways to show up where they are digitally, but the shop experience does not stop at checkout.

It only gets started. And that's what being really efficient about your inventory, about how you're driving that orchestration for that, when that order happens to make sure the consumer, whether they wanted to pick it up in store, get it in a locker, deliver it to them, you know, how do they return it, how do they get customer service, that post-purchase journey is imperative. AI is a huge help in that.

That's where the battle's gonna be won. Um, and that is the key. And where I would say my advice to retailers is look at being efficient and how is your technology giving you this flexibility in an age where the consumer has really high expectations?

Um, and how do you ensure the post-purchase experience is awesome? That's where I'd focus All folks. You heard it here, that retail bar is always being continuously raised.

So if you wanna stay current and relevant, you always gotta kinda, well keep on investing and keep on learning. Hey Mike, thanks for being on the show. Thank you so much, Mike.

All right. ai video series. You can find this episode and others on our website.

We invite you to check them all out. Until then, we'll see you next time.