Agentic Commerce Is Here and It Changes Everything

It's not just a snowboard jacket that I want. It is a snowboard jacket that's Gore-Tex or something similar that has, that's this color that is available with free shipping, that is, you know, it's a really good price. It's all of these things. It's all this specificity. And then me, the user, not going to websites, click, filtered this and the other thing. The AI, the LLM interfacing with the stores through this API, through this universal commerce protocol and then returning back to me what I want. And ultimately, again, one step past that, like, going through the checkout, like, the whole thing. This episode of the e-commerce playbook is brought to you by Attentive. Marketing works best when it feels personal because when people feel understood, they engage. Attentive is the omnichannel marketing platform that helps you do just that by sending personalized SMS email and push messages that customers love. It's marketing made personal. Visit Attentive.com/CTC to see how brands turn customer insights into meaningful moments with Attentive. Hey folks, welcome to the e-commerce playbook podcast. I'm your host Richard Gaffin, Director of Digital Product Strategy here. I comment thread collective and I've joined, or meeting here on a Friday with Tony, the chopper shop, we're talking a little bit about platformer up, but I would, I should say this. The topic that we came into this today with was platform updates. That could be a lot of different things in this modern sort of exciting day and age. So Tony has something that he wants to talk about specifically. We're just going to dive right into it. I'm not going to spoil anything. So I'll have you jump straight in Tony. Richard, happy Friday to you. Like warm here in Southern California. How about Portland? 45 degrees in overcast. The darkness is back, but that's all right. (laughing) Could be worse, I was in-- Could be worse. So this past weekend to do some skiing. And it was one day we were up on the mountain and it was zero degrees, paired with like a 20 mile per hour wind. So it feels like negative 25. I texted my wife and I said, you know, these tropical vacations that you like me to go on are that I've been complaining about 'cause I want to go to the ski trips. Yeah, yeah. I think you might be on to something, actually. Starting to look pretty good at that point. Yeah, man, I feel like that. Anyway, good to be back in Southern California. So happy yard freezing your butt off up there. That's right. So today I want to talk about some nerdy marketing, like technical stuff that I've been, it's been a part of my whole career. And so I'm going to give you the punchline first and give you some of the, you know, Tony's sort of wandering, meandering way of going about it. So we've been hearing about this idea of agentic commerce for some time now. So and Shopify was in the news recently in they developed something in partnership with Google called UCP or the Universal Commerce Protocol. And I think it's really fascinating. And I think it shapes what I believe to be is going to be a meaningful part of the future of e-commerce, okay? So before we get into UCP and what it is and what it means and everything else, it's a little bit of a backstory. I early on in my career, this is like 15 years or so ago, I was working for a small company and then I got this idea about like advertising on the side as a consultant. So I took on some projects. One of the early projects that I took on was this retailer that sold auto parts. And they had three or four million skews. I forget how many, but it was a lot. And the punchline is the partnership didn't last very long. And the reason why is because I didn't know what I didn't know about the data layer. So data feeds and product catalogs and the corresponding ecosystem that existed around these things, platforms like Google Merch and Center. And this was actually before Metas Commerce manager. But anyway, there's this whole like, this whole world that I didn't know anything about. And I went to work in a client partnership that it was really important. And it turned out not to be not very productive for me. But it was a pivotal moment in my career. And it sort of, I didn't want, I learned that hard lesson. And it forced me, it allowed me an opportunity to get really good at this part of the world, this sort of feed stuff. And it's, in many ways, I believe it's coming full circle here 15 years later, where the product data fee and the catalog has always been a really important part of the marketing stack, but it is literally the lifeblood, the DNA of this future commerce world, which you'll hear referred to as agentic commerce, okay? So let's talk a little bit more about agentic commerce and the universal commerce protocol and what it actually is, okay? So the old world user has a thought problem, goes to Google, types in a search query, clicks on a blue link, navigates to a website, browsers, ads to cart, goes into a checkout flow, payment processing, blah, blah, blah, blah, okay. Old world. The new world is that all happens in inside the agents. And this idea of being able to actually transact within GPT or Gemini is what agentic commerce is. And this step that Shopify and Google took in partnership to develop this universal language, this protocol of how that actually works, is a marquee moment in this journey of being able to go to GPT or Gemini and say, I want to buy a snowboard jacket and never actually clicking out to a website. So I'm just curious, I'm like super deep in this and technical and nerdy, but I'd love to get your take on it, Richard, from anything that you've experienced. - Yeah, I mean, I think within my experience so far, like the times that I've used GPT or Gemini or whatever to shop, the initial experience is that it's like obviously a better way to do it. I guess would be kind of like how it would be often say that like the best advertising is or the two factors that are most impactful on sales are price and the recommendation from a friend, right? Or most impactful on whether or not somebody's persuaded and there is definitely a vibe with chat GPT that that recommendation from a friend is essentially sort of baked into it. As opposed to something like, I don't know, wire cutter or whatever, which is, you know, they aggregate different and give you recommendations or whatever, but it still feels like a publication sending it out to many, this feels like it's this one on one individual sort of bespoke shopping experience. And so all that to say, it like it doesn't, it doesn't surprise me and I'm curious to see where it goes, but you would use the metaphor before we jumped on here of the ghost kitchen for this, like I'm curious that that might be helpful to kind of help flesh out a little bit more about what this looks like or what the paradigm shift is here. - Yeah, yeah, I think, I think it feels like a decent metaphor. So ghost kitchen, most people are familiar with it. It's, there's no dining room effectively. Kitchen, transaction. And in the transaction layer here, like happening on an app like Uber Eats or whatever else is the equivalent of sort of this UCP protocol. And then the food gets delivered, but you never go to the restaurant and sit down. And in this metaphor, the restaurant itself, the dining room is the website. And that's the easiest representation and mental model that I've been able to come up with. You mentioned something before about like going into GPT and having sort of a discovery experience for a product. So the thing that I want to emphasize here is the way that GPT or Gemini or any other LLM for that matter gets the information to be able to recommend a product to you is not by us saying, hey, GPT or Gemini, look at us, look at us, look at us. It is a product data feed. And this is the point of how I think this product data feed is it always has been what you actually have to sell. Like if you think about let's use a physical store as a metaphor, there's the physical store, there's the surroundings, the light. But at the end of the day, it's the products in the store. That's what you're going there to buy. - Interesting. - So this product data feed is the products and the process of making sure that that is of high quality and available broadly is what I think the basis, the foundational aspect of what it's going to mean to be successful in agentech commerce. In Shopify, by the way, Shopify is making this a default part of their technology stack. So if you go into the sales channel section of Shopify now, you'll be able to see some of the settings that are available to enable what they're calling their agentech checkout or agentech storefront. And this is the dots that I want you all to connect. This feature in Shopify is what is enabling the LLMs to interact with the product data feed, which therefore enables you Richard to go, it's not that I just want to buy, I'll use my example, it's not just a snowboard jacket that I want, it is a snowboard jacket that's GoreTex or something similar that has, that's this color that is available with free shipping, that is, it's a really good price. It's all of these things, it's all this specificity. And then me, the user, not going to websites, click, filtered this and the other thing, the AI, the LLM interfacing with the stores through this API, through this universal commerce protocol, and then returning back to me what I want. And ultimately, again, one step past that, like going through the checkout, like the whole thing. There's a lot happening in marketing right now, AI, new channels, evolving customer expectations. And if you're like most brands, you're wondering how to keep up while still delivering personalized experiences. That's where Attentive comes in. Attentive is the omnichannel marketing platform that helps brands create messages people welcome across SMS, email and push. Here's how Attentive identifies shoppers across devices and learns from their behavior in real time then AI agents act on that intelligence to personalize every message, send time and channel. The result, messages that feel helpful, not interruptive. Visit Attentive.com/CTC to learn how attentive helps connect people and the brands they love. - Yeah, yeah, yeah. - Right, yeah, I think we talked about this last time, maybe, of eventually what's just going to be happening is robots buying things from other robots or whatever, but at least, at the very least on your behalf. But I think, okay, so then the question this all raises, of course, is like, with this, as a layman, it strikes me, as this is sort, seems unusually advertising proof or difficult to manipulate. And that's obviously like the question that's gonna be on everybody else's mind. How do I like LLM optimize my feed or something like that? And if that's a question of just make sure your feed has like clear names and describes what the product is, that's one thing, but I guess what I'm curious about is like, for instance, to manipulate SEO, you have to write your, I don't know, your recipe blog has to have all these questions. Like, what is eggplant? Where do I find eggplant before it actually gets to the recipe? Whereas in this case, it feels like, those types of like clear manipulative techniques are no longer possible. But tell me what you think about it. - Yeah, I haven't, I haven't been, my initial experience of this whole thing is not advertising proof. I think we're at the stage of this journey where all the major players are working on the user experience. And like the UCP is an example of Google and Shopify getting together to get the technical aspects of how this is gonna work dialed in. Not just the user experience, but, or as a sub component of the user experience, the security, I mean, you mentioned robots, buying things from robots, but I'm pretty like future forward. So like, I'll be like, yeah, here's my credit card, like have fun, maybe it'll be like a secondary credit card, like small limit on it, but. - I mean, dude, if you could give them a credit card and say like, hey, spend, you know, this is my budget for close this month, spend $500, and it just does it and all the clothes are good. What's the problem? - Exactly, this, that is the vision of the future right there. So the, so back to your question though, so the major players are working on what feels like the infrastructure piece of it. And for, for those of us that are advertisers, for those of us that have products, our mandate at this point in time is to understand that the mechanism that by which this happens is a product kept. But to your point about it being sort of, not advertising friendly. Once the user experience gets sorted, the big players are going to monetize the access to the inventory, for sure, 100%. So in this little sidebar here on OpenAI who released their ad product earlier this month, which is, it's all roads lead to advertising, funding, these companies business models. Yeah, just what I'm saying is we have to look fundamentally different because it would be about paid access as opposed to actually, AA actually like running ads to people, I would imagine, maybe I'm wrong about that, I don't know. But at the very least like, I guess I'm trying to figure out like what I'm, what I'm asking here. But it's because the intermediary between your product and the person is this like AI intelligence essentially doing the decision making, the, yeah, what you're having to pay for is to, you're paying for the preference of that bought, I guess, is what the monetization would look like. I see what you're saying, like I understand what you're scratching at. In the, in the old world, we paid for the, the eyeballs like to get in front of the eyeballs. And there is no, there's always been some quality component to the advertising, like the advertising having to be, well, let's do sort of an end around here. So on Meta with like a static or a video ad, for example, the click through how much a user interacts with that ad is a signal to Meta that this ad is high, is good quality. And as a result of that, it's shown more. And the same thing on Google, like all of the, all the tactics that we do, whether it's writing different search ad copy or manipulating the, the images or text in a product listing, and all of it is a mechanism to include, have something good to show a user so that they click on it. And this is, and we're rewarded by the platforms with delivering effective. So I think the interesting, the kind of thing to sort out here is like, if the LLM's are sort of doing that decision making around what, what high quality is, what it means to be the right product for what this person is looking for at any given time. Like how does, how does advertising sort of fit into that model? In prep user preference, I think that's, that's probably the way that it's been done historically. Like not that much different than paid or paid Google advertising versus organic visibility, there's been like preferential access to the impressions if you're willing to. Right, exactly. Yeah, I guess I just like, what shape will like the tactical approach to this take? I guess that's what I'm curious about, because, so, right, right. There's, there, because there's two elements to it. Like you mentioned, like the more straightforward on being like the equivalent being paid search, right? Where you shell out a certain amount, you're spending the right things than Google will show, show your link isn't that, right? And then, but what I'm thinking more about is like what is the SEO of this look like? Because like what are the decision making? What are the factors that influence the decision of the LLM organically? And or is there like, is there going to be no distinction between what's a paid link and what's organic? I mean, at some point legally, maybe we'll have to figure that out. But I guess that's, that's what I'm curious about. How does the LLM decide what to recommend? It could be based on your preference, but there's maybe many products that satisfy that preference. So which one of those is it, how does it determine that? Is it reviews it, like, whatever, like that? That's the kind of thing that I'm, I can't fully articulate, but I'm curious like your thoughts on what, how influencing the decision making capabilities of the LLM is going to look like. Yeah, let me give you a couple things. So we, we have an AI SEO pilot that we're running for several projects right now. And one of the big things that we're, we're focused on in that is to, is to audit and understand the, the state of reviews for the business, aka social proof. And there's some places that are turning out to be really important to the LLM's, like, conversations on Reddit, but that's not the only place. What it appears that it appears that the LLM's are looking for some broad signal about the validity of a brand, the quality of a brand, and they're, and there seem to be looking at that from, with a, with a, with a broad view, that's, that's one, you mentioned preference. So this, the preference thing connects back to the product data feed. So the more, I mean, this has always been true of product data, the more detail, the more, and the more descriptive you are, the more likely you are going to meet somebody's really unique preference. I think the, the third piece of this puzzle back to this UCP universal commerce protocol and sort of LLM's interacting with Shopify stores, one of the things that we've heard some conversation around is the LLM's sort of being able to negotiate around, is there any offers? Is there any coupon codes for this product? So I would bucket this all into price, right? So if the, the price is competitive, especially if it's a retail product. And if there's some sort of mechanism in, in that process that's going to make the user experience better, like, oh, I got the best deal on it, then that is for sure something that's going to be part of this, part of this ecosystem. Right. No, I mean, that makes sense. I mean, ultimately, just, like many of these things feels like it comes back to just make a great product that people will talk about and they talk about in certain places and that, that will be what the LLM is sort of trilling for. But inevitably, there's going to be some sort of, as marketers, I think that makes you feel kind of a little bit out of control. So there's going to be some sort of element of, like, the SEO of LLM or, I guess. It's no more control than it ever has, by the, yeah, marketing and in this, whether it's on Meta or Google ads or Super Bowl commercial, that, that end note has always been a distribution mechanism, a way to get a message out about what it is that you're selling. And so product and brand has always been like a critical factor in, in the marketing world, especially, you know, specifically here at CDC, we get really methodical about how to understand the unit economics of advertising and, et cetera, et cetera. But it doesn't supersede having a great product to sell in a good story to tell that people resonate with. Right. Of course. Yeah. I guess everything in like an instance is where, well, I mean, yeah, having a great product to sell, but not a great story to tell. I think is an issue people kind of run into or and kind of like the whole point behind something like SEOs. You may have a great story to tell, but there's certain ways that you need to tell it in order to get the algorithm to actually read it, right? And so that, that's kind of what I'm, I'm thinking more of like, what are the techniques that you need to tell a story with in order to trigger the LLM into thinking that like, oh, this is something interesting. And maybe at this point, because it's way it thinks is so complex, all you can really do is like make a great product and have people write good reviews about it. But the jury is still out, right? Yeah. But right now, today, a recommendation that I would have for anyone listening to this is number one, go do some research on this UCP on this protocol that's been developed between Shopify and well, because it's going to underpin this infrastructure and go do some research into Shopify's agent check out because that's literally the piece that sits underneath this thing as well. And the more that we all familiarize ourselves with what's happening sort of under the hood, I think it's going to get us all to that next point that you're talking about is like, how do we, how do we, how do we behave in this new world? Right. Yeah. So this is, this is still something that's cooking. We're not ready quite yet to figure out how to do it strategically. Yeah. I've spent the first, I spent 30 minutes this morning trying to get it to actually work. So I was like in Gemini and I was like, Hey, I want to buy, I want to buy snowboard pants. And I'd like you to do the checkout for me and to and it sort of, it, it didn't work. Let's just leave it that went just like a couple of different loops. So they just started talking about it or the sponsor, I was like, yeah, it to me now, but I think that would be like a third piece of the puzzle is to is this kind of one of my favorite ways to do advertising is just to experience advertising. And so I encourage everyone listening to try to try to get these experiences to trigger in their daily driver, whether that's Gemini or GPT or whatever. You can also preview it and Shopify. If you go to Shopify into the sales channel and go into the agenda, check out, you actually see a preview of what, what it's expected to look like in some of these tools. But I want to, I want to like sort of shift gears just a little bit and hit one more connected idea that I think is pretty cool. So I was up at the meta office earlier this week to do some filming for their performance talk series, which was really fun. I got a chance to spend some time with our agency partner who shared spilled some of the tea on the things that are important for for meta this coming year. And he is funny, funny enough. He's like product feeds. I was like, oh. And so the, it all kind of clicked when I was up there, I'm not, are you a TikTok user at all? No, I've been off for a couple of years, it's been great. But continue. Well, so TikTok, yeah, I'm not super big user either. I use it to experience advertising products. Sure. They have a new product that is, that's pretty cool. And it is like a hoverable and clickable thing that exists on a product that's in a video. So if you're watching a video and there's like a shirt and the, you see a product you like, you can click on, right? This is all being driven by product data feeds. And so what I heard from our folks at meta was everything is going to be shoppable this year. So I think what we can expect to see is a similar functionality in a real specifically where if meta's algorithm is able to identify a product in a real, there is very likely to be a clickable interface and potentially an agent to check out all with it. There's one more thing that the audience can go and research a little bit. It's called SAM. It's a model that meta has been building called the segment anything model. So SAM 3. So you can go and read about this. And essentially the easiest way to understand it is it is able to analyze a video and isolate any object or subject in that video from the rest of the content. So think about this. I'm making a, I'm in a video. I'm wearing something, right? The SAM algorithm is able to identify this object. It's able to connect it to a product data feed for the retailer. And then a clickable link go by this thing. Crazy. Crazy. Wow. Yeah. I mean, yeah. It's a brave new world we live in for sure. I mean, just thinking about like the possibility of, so I was thinking back to that thing that I was saying before about like, hey, can I give Gemini my credit card and say you have 500 bucks to spend and give me this, whatever. And just thinking about like, yeah, I saw some guy on a TV show and I really liked what he was wearing could, could there be like some scenario where or like I'm watching a YouTube clip of it. I'm like, just give me whatever's in there or give me something like it or whatever. And I don't know. Yeah. It's a, it's just kind of mind blowing, I guess. I mean, it's really cool. If you think about it, like we've been in the UGC content creator era for a long time now and I, and that, that style of content is absolutely crushing on social advertising. Well, it's, we have all sorts of data points in the agency about how, how important this is for our brands to make this talent content. In some ways, it is, it's still sort of, it's unequivocally advertorial, right? It's a person that's advertising a product and so it's going to continue to be important. But you can almost see this, this future world where you're watching something on TikTok or you're watching something on a real on meta or YouTube for that matter. And nobody is intentionally advertising anything. They are just existing and probably being cool in whatever thing they're doing. And the infrastructure underneath all this is like, that's a product. And I'm the user going like, that guy did a cool snowboard trick, but actually as Jack it looks rad, click. Yeah. Yeah, yeah. Exactly. Yeah. That's fascinating. Oh, just be like, hey, send that, ship that to my house. Yeah, yeah, yeah. I put it all in. Exactly. Just speak it into existence. All right. Well, so I mean, I feel like the takeaways from this, practically speaking, they're just like, just do some research, get prepared. Any kind of other like sort of like practical takeaways you want to pull out of this here? Yeah. Yeah, I mean, so you sent this, but go, go read about UCP, go look at, educate yourself on the agent check out in Shopify, learn about the Sam algorithm on meta. If you don't know what a product data feed is, or you don't understand what what it means to have good high quality product data, you need to educate yourself on this because this is going to be a material part of the future of commerce. And if anyone buys any of these $60 or CPM text ads on OpenAI, hit me up because I'm curious about them. So apparently, no, you know, conversion tracking, you just clicks and views only, I guess through text ads at the bottom of the search query. So it's sort of like, whoa, really seems pretty expensive, but interesting. All right. Hit up your boy if you buy some OpenAI ads because I'm curious. All right. That's good. All right. Anyway, product data feeds get fluent on it now because it's going to become really important in the future. All right. I think that's going to do it for us for this week, folks. Thanks for listening, everybody. We'll see you next time.