Latest / Retail Media Breakfast Club / What is a 'CommerceGPT'? ['Snips' from Cocktails & Commerce]
Transcript
- What is a CommerceGPT? === [00:00:00] Kiri Masters: Hi, it's Kiri and today is another Snips episode where I share a snippet of a conversation from one of my favorite podcasts in the commerce space. And today's snip is from one of my longtime favorites called Cocktails and Commerce, which is hosted [00:00:15] by Brian Walker and Bill Friend, who are two longtime commerce leaders who talk shop with industry operators and thinkers over a drink. [00:00:24] I was on the podcast a few months ago when I was in Thailand, bright and early, had a mocktail. It was a [00:00:30] great conversation and what I like about it is there's a lot of talk about commerce tech, which I am sort of, I feel like I'm on the periphery of, and it's an area that I wanna understand more. So I really appreciate their ability to talk to, um.[00:00:45] [00:00:45] Commerce technology leaders, but also keep it at a level that, uh, an everyday pleb like me can understand. [00:00:52] [00:00:53] Kiri Masters: So in last week's episode, they were joined by John Andrews, [00:00:57] the co-founder and CEO of [00:01:00] Cimulate a generative AI company based on commerce platforms. And I'd been hearing from Brian and. Bill's commentary on their excellent substack about a concept called Commerce GPTs. But candidly, I [00:01:15] hadn't really taken the time to get fully up to speed on it, and I, it's been on my catch up list. [00:01:20] So I was really pleased that on this episode, they dug into that concept in a little bit more detail. [00:01:26] We're gonna jump in at the point where the hosts have just asked John to [00:01:30] explain what a commerce GPT is. [00:01:33] John: And I think commerce, GPT, it's frankly a nod to the new era of the commerce that we're, we're in and, and a new element within the stack that's going to significantly change how that experience is delivered. Listen, early [00:01:45] on we used the term customer, GPT, as the thing to kind of describe our engine and then seeing what you guys were bringing out into the market. [00:01:53] And this market definitely is a, a commerce GPT market, right? And so we have the, we've adopted it, we have the simulate [00:02:00] commerce, GPT, right? So just to keep it simple, I think it needs to account for two things. You know, like you said, what does this actually mean and what does it, what needs to take into account? [00:02:07] I think two things. One is, go back to what I said earlier about understanding customer behavior, okay? And this is that sequence [00:02:15] model. This is understanding that customer journey and making very efficient use. Of customer behavior to understand what the right next best thing is to put in front of that customer at any point in time. [00:02:28] Right? So that's, that's one. [00:02:30] I think the second thing it needs to do is it needs to be a model that, a model that understands effectively everything that the internet knows. And you can think about this as everything that's ever been written on a particular topic or about a [00:02:45] particular product, and use that to drive a significantly better customer experience. [00:02:50] So if you could do that, think about chat GBT, right? They've spent hundreds of billions of dollars basically sucking up the entire internet and it knows a lot [00:03:00] about everything. Imagine if you could have something very focused that knows. Absolutely everything about a particular market area or product area for a particular retailer and everything about that Retailer's, customers and their behavior [00:03:15] and how they talk about certain products, how they make decisions about those products, what they like about it, the language that they use. [00:03:21] And you use that model, uh, that commerce GPT to drive the customer experience. That's kind of going back to Kai and, and my description to him. [00:03:30] Think of that as the brain inside that guy that was selling him and, and getting him all pumped up about the new Nintendo two switch Did you know that leading retail media networks [00:03:45] drive 85% of their ads through mid and long tail advertisers? [00:03:51] Kiri Masters: Miracle Ads provides full funnel ad formats tailored to both one P and three P advertisers leveraging unique AI [00:04:00] capabilities that provide unprecedented levels of relevance and engagement. Retailers who want to capture ad spend from the long tail of three P Marketplace sellers use miracle ads in their tech stack. [00:04:14] Learn [00:04:15] more@miracle.com. That's M-I-R-A-K l.com. [00:04:20] [00:04:21] Kiri Masters: so if a commerce GPT is the brain that connects data, behavior, and decision making, the next question is, where [00:04:30] does it operate? Because increasingly people aren't just engaging inside a retailer's own ecosystem. Answer engines like Chatt, PT, and perplexity are becoming new gateways to discovery places where shoppers ask questions and where product [00:04:45] visibility will increasingly depend on how accessible your data is to these AI systems. [00:04:50] And that's when John was asked by the hosts how brands and retailers can position themselves for visibility in that world. [00:04:56] Bill: I wanted to ask you a big topic and that's around the [00:05:00] impact of the answer engines. There's a lot to think about here, but let's just start with discoverability through the answer engines, specifically around shopping and products. [00:05:08] Many focus on extended product information feeds and SEO like tactics to drive discoverability on the [00:05:15] an answer engines in the market today. But I wanna hear what you expect in terms of how this may work over the long haul, and what's the role of commerce GPT in that future? [00:05:24] John: This space is moving incredibly fast. [00:05:28] There's a lot of unknowns here. I feel [00:05:30] like there's a lot of snake oil out in the marketplace in terms of people saying they know exactly what's going on. But even the people who are driving this, right, like open AI and perplexity, they don't even know, right? They're trying to figure it out and, and let's be clear, based on their [00:05:45] market caps and the amount of money they've raised, they need revenue and they need some of that revenue to come from commerce. [00:05:52] Now, whether that comes from selling ads or actually selling products, I think with checkout we're gonna find out. We're certainly gonna find out, [00:06:00] but like they're going to wanna push as far as they can. I think it's interesting. I had a conversation today with a Gartner analyst on this topic, right? And this analyst actually released a paper recently that Gartner is calling within their evolving insights topics, [00:06:15] where they're actually editing these documents. [00:06:18] Real time as things evolve and they're showing the edits that they're making. And so like on one hand, I, I give them a lot of, I give them a lot of credit for the transparency here because it's not like you can say like, this is the answer and this is [00:06:30] gonna hold for the next six months. It's not gonna hold for the next six minutes. [00:06:33] What's my take on this and what's simulates take on this? I think first off, from a brand perspective, I think brands need to do whatever they can to get visibility into these answer engines. I think things are moving quickly. I think the [00:06:45] level of the amount of traffic that they're gonna begin to see coming from Perplexity and Chat, GBT, I think is the two big ones that we are really seeing right now. [00:06:54] Like bringing some traffic in is gonna be increasing and they need to learn from every one of those interactions, [00:07:00] right? They need to, they need to test it out. They need to learn. And so if that means handing over their product catalog, so be it. Let's let chat, GPT, perplexity, et cetera, understand what products they have and that they are, they are the experts [00:07:15] within whatever space, right? [00:07:16] So going back to West Marine, if somebody on Perplexity asks something about boating. Well, they better be damn sure that, you know, west Marine is in that, uh, that solution set in terms of, uh, sites and product catalogs [00:07:30] that matter to that, to answering that question. I think longer term, something like an MCP server is going to be a bigger part of the, it's gonna be a bigger part of the infrastructure with commerce. [00:07:41] GPT being the key tool on the server [00:07:45] that's managing kind of that agent to agent interactions. Nobody was talking about CPS six months ago, four months ago. Right. Everybody's talking about them now. MCP is a, it's a protocol, it's a fairly thin layer that kind of sets that [00:08:00] protocol. What is important is what are the tools that you have available on that MCP. [00:08:05] To serve whatever task you have at hand. And for commerce, when there's a question about products, we believe that tool needs to be the commerce GPT from [00:08:15] West Marine as an example. Okay. And that when perplexity comes, they can come to the West Marine MCP, get a very rich answer. And then ideally that gives that confidence or perplexity to include it in one of the tiles on, in their answer in terms of what the [00:08:30] customer sees at the end. [00:08:30] And then West Marine gets a sale out of that. And [00:08:33] Brian: it'll have to be fast [00:08:35] John: as well. It's gonna have to be lightning fast. And that's why, you know, a commerce GPT doing something that runs through a open AI or kind of [00:08:45] does a lap through a more, kind of a larger engine, uh, after, you know, five or six seconds, a few seconds, that's not gonna, that's not gonna fly. [00:08:52] It needs to be 50 milliseconds. Right. Bringing that response back to that large kind of foundation model to incorporate it. [00:08:59] Kiri Masters: That [00:09:00] vision of agent to agent commerce where AI systems directly talk to each other, sounds like science fiction, but many people in the space believe that it's coming down the pipe pretty fast. Still, that shift won't happen overnight for [00:09:15] consumers. It's a new level of trust that they need to have. [00:09:18] Letting an AI agent make or even complete a purchase on their behalf. And this is where I was glad to hear Brian's perspective on how he thinks this will play out and what advice he'd give to brands [00:09:30] who don't wanna be left behind. [00:09:31] Brian: Yeah, I think the consumer sort of behavior change and trust and so forth that is gonna be required for this to become a really, really significant like percentage of commerce. Yeah, I think that may take time to play out. [00:09:42] There will be early adopters, there will be [00:09:45] categories where this makes a lot of sense. There will be consumers who are really excited about it and are actually willing to live with some of the lumps that might come along the way. Recognizing that this is early days. We also see some published research from the likes of Anthropic and [00:10:00] OpenAI, talking about agents, uh, sometimes being manipulative and or, you know, depending on their incentives, like kinda steering answers and so forth. [00:10:09] Not just being syco offense, but even being a little manipulative. Not to suggest that [00:10:15] that's really gonna happen here, but it's an indication even at the foundation models, levels today of some challenges that may lead to some. Unfortunate, viral, unfortunate events and other things that are just gonna play on people's minds. [00:10:28] The technology issues [00:10:30] will, I think, get solved and that may happen in the relatively near future, but I think the customer behavior changes and sort of trust around what is, feels like a very different process where I'm giving over some control. Might take longer to play out. You [00:10:45] know, in our forecast when we project five years out, you know, we can be kind of bullish. [00:10:48] But anything short of that, I start feeling a little more conservative and thinking about how fast this may ultimately play out. Yet, I think brands and retailers and B2B companies, for that matter, [00:11:00] need to start preparing and in a sense, have some of the foundational pieces ready and potentially be able to experiment early on in order to learn. [00:11:09] Recognizing it may take time for this to really become a significant percentage of their revenue. [00:11:14] Kiri Masters: And I [00:11:15] couldn't agree more with that final point from Brian. The smartest move right now isn't perfection. It's participation, the brands who test, learn, and adapt early, will already have a seat at the table when this behavior starts to go mainstream. [00:11:30] Thanks for tuning in and I'll catch you tomorrow. [00:11:32]