← Insights Unlocked
Insights Unlocked
Who owns the customer when AI does the shopping?
00:27:02
Cordial's Matt Howland on the Shopper Context Protocol, AI agent commerce, and how brands can keep customer context as AI agents start shopping.
Loading summary
Transcript31 lines
- [00:02]
Nathan Isaacs
Welcome back to Insights Unlocked. In this episode, I sit down with Matt Howlin, President and Chief Product and Engineering Officer at Cordial. He also chairs the working group behind the Shopper Context Protocol. We get into what breaks when AI agents start shopping on our behalf, why brands risk losing their customer relationship entirely, and what leaders can do right now to prepare. Enjoy the show.
- [00:28]
Podcast Host / UserTesting Narrator
Welcome to Insights Unlocked, an original podcast from User Testing, where we bring you candid conversations and stories with the thinkers, doers and builders behind some of the most successful digital products and experiences in the world, from concept to execution.
- [00:47]
Nathan Isaacs
Welcome to the Insights Unlocked podcast. I'm Nathan Isaacs, Principal Content marketing manager at UserTesting, and our guest today is Matt Howland. Matt is the Chief Product and Engineering Officer at Korjul, where he leads product innovation at the intersection of AI data and customer engagement. He's also the chair of the Shopper Context Protocol Working Group, helping shape how brands and AI agents interact in the future of commerce. Welcome to the show, Matt.
- [01:15]
Matt Howland
Thanks, Anthony.
- [01:16]
Nathan Isaacs
It's great to be here, Matt. Before we dive more into talking about the Shopper Context Protocol, I'd love to start with your journey. Can you share some of your background? What experiences led you to the work you're doing today at Cordial?
- [01:30]
Matt Howland
For sure. I can go way back when I was about 7 years old, that's when I started programming. And it was because I really wanted my computer to talk to me. Maybe I need more friends, I don't know. But that was. That was nickel. And, you know, fast forward to today and guess what? Computers are talking to you. But between then and now, I spent the bulk of my career in Silicon Valley, kind of going from the Martech sphere and kind of being a vendor side, I ran a large loyalty company that powered the loyalty programs for Targets, Kohl's and the like, and then through the actual retailer side. So I helped George Zimmer start Generation Tux. I did a little bit of work at Private Equity, overseeing a couple brands there and then, now where I am at Corjo. So I've kind of seen both sides of the equation being the vendor as well as trying to actually drive that engagement with customers as a retailer. So I've kind of seen both of those and I really saw this need for context in the world of agents and in the world of LLMs, that context of that brand and customer relationship.
- [02:38]
Nathan Isaacs
For listeners who are just hearing about Shopper Context Protocol for the first time, how would you explain what it is and why it matters right now?
- [02:47]
Matt Howland
For sure. So Shopper context protocol was born out of the need for brands to maintain a relationship with their shoppers through agentic commerce or through chat apps like ChatGPT or Claude. And it kind of originated when ACP Agentic commerce protocol came out. That was with OpenAI and Shopify and some others at a consortium to make it so you could purchase my flow chatgpt. And I thought that was awesome. I thought I'd done a decent amount of discovery there. It made a ton of sense. And when I looked at the protocol, it was really, what are your products? Here's a cart and here's a checkout. And it completely disintermediated the brand themselves. So you had no agency over how anything was merchandised or anything like that. So that's when I kind of said, I think this is a solvable problem. I think we can make a discoverable secure protocol that it really pulls in that customer context from a brand lens. And at cordial, we've done this for a long time, but using that context to send emails or texts or whatnot, this we just looked at as another channel, but an interesting bidirectional channel. So that's kind of where the idea of it came got to an MVP in a week that ran on the rails of mcp, my context protocol. So it sits alongside acp. It's not really meant to replace it, it's meant to sit alongside it. And then nrf, Google introduced something called UCP Unified Commerce Protocol. And Unified Commerce Protocol and acp, they kind of compete with each other. They're both trying to get the rails for that transaction and that cart. In my opinion, UCP is much better thought out and it's evolving more quickly. So it actually has a place to get loyalty information as a place to get some offers. It keeps expanding and if I'm honest, it would actually be great for UCP or a standard like that to evolve, to consume a lot of the things that SCP has today. Because I designed as an open source protocol quite intently, because I wasn't trying to design a product to serve this need. I was really kind of saying, hey, as a consumer, like if, if I shop boot bar all the time and I'm shopping for some boots for the Houston Rodeo, I don't want you to go and serve me something somewhere where I've never shopped before.
- [05:09]
Nathan Isaacs
The and as I just think about this, we've had these sort of conversations over the decades, right, with any new technology and we start thinking about, you know, whatever, because I'm not a technical person, I'm just a marketer, but I, you know, like, I know we've done this, right, We've done it on, you know, you know, the meta stuff that goes into SEO and, and, and all that, right. It's, it's just we're going so fast, we have to, you know, think about these things. And it sounds like that's kind of what happened here is like 100%.
- [05:47]
Matt Howland
And I think you also have this interesting thing where, and this happened with meta and others where there's a tension between platforms that want to create an ad ecosystem. So if you actually look what's in their self interest, it is actually to disintermediate that relationship because then you can own the payment rails and you can charge an extra percent on that and you can charge for access to those customers. Right. So as you kind of looked at these evolve, you know, this has been Google's business model for the last two decades. So it's kind of the reins on the wall on where that's going. And with ChatGPT introducing ads as well, you kind of saw this as an initiative to really, truly disseminate the customer and it's kind of pay for play. What I tried to do is go to the retailers, some of our customers, and say, hey, I think we need a voice in this too. Because if this is just another walled garden, that's not going to be great for anyone. Because you can't own your customer's relationship. You can't understand the intent. Um, that was one of the, one of the interesting things about the protocol is it does bidirectional communications on the concept of intent. So if you think about, you know, doing some product discovery on ChatGPT, shopping for boots or something like that, it gets a lot more information than just clicking on a website. Like it's going to know that I'm shopping for boots to go to the Houston Rodeo, you know, in March or something like that. And that intent should actually be durable and transferable back to the brand so that when you end up on the brand site, they, you know, the customer isn't having to reiterate themselves through clicks and browsing. So that was one of the pieces we had to SCP that really kind of try and drive that intent sharing. So whether you're in store, on site or in ChatGPT, the, the merchant has an understanding of, you know, why you're shopping with them.
- [07:34]
Nathan Isaacs
And in the end, I mean, everyone's sort of being served here, right? The, the Googles and open AIs of the world. They're doing it for their reasons, you guys. And the retailers are doing it for their reasons. In the end, the customer just has a better experience if it gets this sort of information when they're shopping. Right. I don't want to go through the process of having to filter, you know, through, you know, I need to go to the fat boy men's section rather than, you know, women's or whatever it might be. It just makes it smoother for me to buy something in the little time that I have.
- [08:11]
Matt Howland
Right, that's right, it makes it smoother. And also, you know, the, the concepts like loyalty, which exist for a reason. You know, I spent a good, good time and good amount of time in that space. You get to carry those with you. So, you know, as you're looking at different merchants being discovered through, you know, agentic interface, you can kind of choose one of those. So I think that's really important. I kind of say that, you know, the way it's evolving right now is how you shop on Amazon and you shop for, I don't know, a shoe rack or something generic. And there's literally 10,000 of these things and they're undifferentiated, as is your customer relationship with those different brands. We don't want to recreate that in an agentic world. I think the agentic world can learn so much more about you. And if it also includes the brand in that learning, it's just going to be a better experience for the customer and.
- [09:07]
Nathan Isaacs
Oh, yeah, absolutely. Right. And I mean, that's one of the challenges I have now as a consumer is there are brands I want to buy from that I trust. And then they're all the sort of like, get rich scheme, you know, brands that are on Amazon that are trying to sell me that product and I'm like, do I really want to buy, you know, food or medicine on Amazon? That.
- [09:32]
Matt Howland
That's right.
- [09:33]
Nathan Isaacs
Where's that coming from? You know, these are real thoughts I have occasionally. The 100 and, you know, so what's that risk if. If for retailers, for brands, if they don't engage with standards like scp, you know, is this the end of like those sort of brand loyalty programs?
- [09:56]
Matt Howland
You know, I mean, we did a study at Cordial over the holidays to kind of see how many customers were actually using. And I'm using Gentex loosely here, but the concept of a ChatGPT or a cloud when they're actually purchasing, and about 65% of them use them for some element of discovery. This was before checkout was there. So it was really kind of clicking out there and going to the link. But if all of a sudden you take 65% of those users that are shopping and you have the totality of the experience there, the product surfacing, any merchandising, the checkout itself in fulfillment, you know, you're completely disintermediated as a brand. So the brand at that point is ChatGPT, right? And they're the ones who have the context to your customer. They're the ones who understand, you know, oh, hey, I tried this size and it didn't fit quite right. You don't get to consume any of that. So that's one thing is you're just going to have less fidelity of your customer that all that information you could capture is going to be gone. The second one is over time is when they become another walled garden that's going to become pay to play. So right now, you know, the mechanics for a retailer is you optimize your site for something like aeo, so you put, you know, Q and A on there, you create a robust product file that really can help service your, your products. But at the end of the day, what it's going to come down to is who's going to pay me the most to show their products first, right? That's, that's the direction you were headed. And if that's where we end up, it's just yet another walled garden where your brand is completely dismutated and it's just, you know, another tax you have to pay as a retailer.
- [11:33]
Nathan Isaacs
So where do we go, where do we go from there? Because as you're telling me that, I'm just like, yes, we went through this in the infancy of all the social media, right? We had all that free organic reach and then all of a sudden, well, no, you don't get that reach anymore. It's like a sliver of what you did. Unless you're paying for paid promotions and stuff like that. And I'm speaking as a marketer and talking to reaching my audience. But so what, what happens? What's the future? How do we. What can be done?
- [12:04]
Matt Howland
Yeah, for sure. Well, I'm an optimist here, so, you know, I brought s piece of the table. Not really thinking that it was going to become the established protocol, but really introducing a concept to the market that hopefully others like ucp, others can adopt. UCP has already pulled in some of the loyalty components, so that piece is starting to be surfaced. They've extended their, some of their customer information, they get more fidelity there. So I think that the. Some of the protocols are starting to go that direction and I think the reason they're going that way is as ChatGPT and some of these others have tried to roll out monetization. It's one. I think it's very difficult. It's not as easy as search because it looks incredibly biased. It looks like ChatGPT is becoming the salesperson. And I don't think that's the relationship people want with these models today. I think people want a relationship with these models where it's surfacing information, making it easier, but not sewing on behalf of. So I'm an optimist that I think customer behavior and customer interaction with those ads is going to lead to where those ads are really informed by that brand customer relationship. And I think that is the direction it's heading. If you kind of look at the way the protocols are skewing and even look at the way that, you know, some of the ad testing skewing, it's heading that direction. I do think that, you know, the more that brands could get involved with, you know, their spending on these ad platforms and whatnot and their access to data, the more they can steer those. So I think it's really kind of the action there. But this is one of those ones where I think the user interaction is going to drive a little bit of a different result than it did in social media. At least I'm hopeful.
- [13:48]
Nathan Isaacs
Yeah, it would be interesting. One of my colleagues, Mike Mace, as you were talking there, I reminded of a study he did going out and asking people to do some research on dishwashers or something like that, using their chat, using whatever LLM they use, and a lot of them. And then. But what he found in that research was that they did not want to use the chat to buy the dishwasher, but they wanted it to inform which one should I look at that makes less noise or something like that. Right. In the end they want to go to, you know, I forget the name of my dishwasher, but GE or LG or whatever it might be and I want to go to that website and I want to look about their details. And so it was informed from that. And so I think that kind of says. Gets to what you're saying as well is like, in the end we still want to buy from the brands as the shoppers, right? As the buyers, I think so I
- [14:53]
Matt Howland
think we still see the, the brands as the, the source of trust. Right. I think, you know, LLMs have gotten so much better over the last year. But everyone's had their moment where they use an LLM and you see something, you're just like, no, that's wrong, that's wrong. And, and I think that has shaped some of our interactions with them where we kind of see it as that helpful buddy, but probably not the authoritative source. So someone who kind of goes, hey, you should take a look at these. You're like, great, but you can hold off on the details and really kind of verify some of that. I do think some of the stuff that Gemini has come out with grounding, obviously given their prevalence in search, is making that a lot better, so it is more trustworthy. But I mean, you even look at some of those search results as they put the AI block up top, some are pretty bad. So I think you're, I think the user's still being trained, like, listen but verify. And I think because of that, really, that brand source is still important there. And I do think it also helps, you know, that discovery motion really kind of scope it down, but then have exploring. I also think there's another piece to, you know, retail in general in that I think retail is largely a recreational activity. Some of it is done because you need shampoo or you need toothpaste or either rusty shampoo, we both need toothpaste. You know, you can kind of go down that direction. But I still think people like to shop. I still think people like to like that, you know, finding that, that, that prized possession and a discount place somewhere, like it's still, you know, it drives some emotion. So I still think that exists too. So I think there will always kind of be this modality where, yeah, let me kind of hone in on what I'm looking for here, but let me go do a little bit of, you know, searching and recovery over on the side of the actual brand website.
- [16:49]
Nathan Isaacs
Interesting. Yeah, definitely. I'm a big shopper of Shampoo. The. For people who are listening, Matt and I are both follically challenged, so.
- [17:02]
Matt Howland
That's right.
- [17:05]
Nathan Isaacs
You know, I, I, I, I'm remiss. I didn't ask more about Cordial and I just was wondering, can you share us, you know, a bigger picture of what you guys are doing at Cordial? What's the company's focused on and, you know, how's that making the biggest impact for brands?
- [17:23]
Matt Howland
For sure, for sure. So at Cordial. So Cordial was founded about a decade ago and the concept was send a better message. The concept was really how can we better engage customers on behalf of retailers to really engage with the brand. And that comes from really understanding the customer, understand the brand and understanding the product with which they're selling. So over the years, Cordial has really evolved into, at the end of the day, the customer context engine that's really powering email, sms, push notifications for some of the best brands in the world. So we work with brands like Levi's and Abercrombie and, you know, Tapestry and countless other, you know, great brands to really not only message their customers, so we do power all of that, but more importantly, understand their customers. So we're not treating them all the same. We understand why one customer might be shopping at Boot Barn because, know, like I said, I went to the Houston Rodeo last year, I needed some boots. That might be one use case, but there's plenty of other folks who are coming in there and saying, hey, I need some new workwear or some new jeans to wear to work tomorrow. And really understanding the difference between those two shoppers, me trying to kind of, you know, wear an outfit for the Houston Rodeo versus someone who's, you know, really trying to kind of fulfill their need for workwear. Understanding those two user intents or customer intents and then communicating to them through whichever channel they choose, you know, whether it's in store or, you know, email or whatnot, really communicating with them in that manner is what we do really well. And it's, it's been really fun to do this as AI has emerged because the term I like to use is uplink. So it used to be you learned about a customer by what they searched for, what they clicked on and what they purchased. And that was like, it was like a 28, 8 odd modem, right? You kind of got a blip here and there. With LLMs and especially with people that have deployed on site chat and things like that, it's like all of a sudden having a fiber uplink, right? So we can take in all that unstructured data and understand, oh, hey, Matt's going to Houston in March, here's the things he's shopping for now. And then we can suggest some other things that would go with that. We can really understand the totality of, you know, what that customer is trying to do. So it's been really fun kind of being able to go into the unstructured side of things and the same thing on understanding products. So similar to the example I gave before, you know, the jeans or, excuse me, the boots, someone's going to buy to work in an auto shop versus a boot someone's going to buy to go to a rodeo are likely fairly different and the attributes expressed on those are likely fairly different as well. So being able to do the attribute enrichment to really understand that clustering of intent, something we've gotten pretty deep in and is linking this back to the AI commentary. This is one of the primary drivers for really successful AI. AEO as well is really truly understanding the intent that people might be discussing with an LLM around that product. So surfing some of those product attributes there is very important as well.
- [20:38]
Nathan Isaacs
I'm kind of speechless only in the sense of like, wow, this is, this is everything that's on the horizon for us. You know, it's, it's, it's thinking about it and, and just the speed and, and I'm reminded of LinkedIn post you had here in the last few months where you just talked about your aha moment. You called it something else. But it, you know, as other listeners are listening to the show, as leaders are listening to the show and they start thinking about fitting AI into their organizations if they're not already, if they're, if they are retailers thinking about this and how they are, how commerce is going to be in the future, what are, what are things that they can start thinking about and preparing for, like SCP within their own organizations?
- [21:34]
Matt Howland
Yeah, for sure. I think there's kind of two pieces of advice I'd have there and one relates to SAP, another relates to we just did an enterprise wide deployment of anthropic and really changing our organization to use AI internally. So we're very in SCP and things like that. And I would actually say this holistically across AI, jump in. So so many of these things are low friction to try and it doesn't necessarily have to be trying to solve a problem that's exactly related to your work or something there. One of the things I've done with our team is tried to explore a breadth of ideas that can be solved with AI. Some of them business related, some of them fun and personal. Because what I found time and time again is if you can help someone with something that sparks their curiosity, where they keep going and going and going, that's when they get that aha. And they're like, oh, this is what I can do with that. And then as soon as you kind of have that spark, it just kind of, it just kind of takes over and people start trying different things. As you try different things, you understand what works and what doesn't. Because there's a lot of stuff that doesn't work. And there's a lot of stuff that works really well for the internal deployments that still holds true and that jump in and try things. But one of the things I found really interesting is some of the people who have been most successful are the people who cross boundaries in the organization. So the people who are just optimizing just product or just engineering or just ce, they're doing great, you know, but the people who are crossing that boundary between product and ce, so they see a little bit of the problem in product, a little bit of the problem in ce, they're force multipliers because they're seeing problems that no one else has seen. So I started kind of trying to put some framing around this and the idea of like how do I intentionally create these discovery surface engines or surface areas. So if you think about AI and if you think about what AI can be though, that can take over. That's the jobs they would like to find. So if I have to respond to an RFP or you know, if I have to, you know, reconstruct a roadmap or things like that, there's just some jobs that AI does pretty well and it takes what would have been a few days down to minutes. So if all of a sudden you have time to distribute elsewhere, what do you do with that time? And we're leaning very aggressively in how do we discover the next problem to be solved. That really adds value to cordial. And so I think if you kind of lean into this idea of, okay, let's give the AI agents or agentic kind of processes as much of this well defined work as you can not so that you can change your headcount or get cost. That's crazy. What you want to do is discover more and more things that can be done to add value for your customers and for your own business. And I think that's the framing that's really worked for us is give the mundane stuff that you don't want to do to something else and make sure you've got evals and boundaries and all the good hygiene around AI process there. But really lean into, okay, cool, so that's done. So how can I actually drive more value for this business and for our customers?
- [24:45]
Nathan Isaacs
Awesome. I'm hearing a lot of people have that sort of framework there, you know, not. And I'm not saying in a very smart way. Right, that's the right way to, you know, if you're just doing this to race to the bottom on costs, you're never going to win in long term. Matt, I appreciate you coming onto the show and I enjoyed our conversation. How's someone learned more about you, your thought leadership, how they learn about Cordial and the latest on Shopper Context Protocol?
- [25:16]
Matt Howland
Absolutely. The easiest way is connect to me on LinkedIn. It's just Matt Howland. I kind of spawn everything out of there. If you go to cordial, it's www.cordial.com. we can kind of explain what we do there. And then Shopper Context Protocol is shoppercontextprotocol IO and if you're really interested there, we do have a working group. There'll be a form there to sign up. We meet quarterly, kind of try and digest, you know, some of the newer things going on with agents, long live agents, and things like that and try and solve them. But, you know, for people that are thinking about any of these problems, I'd love to just, you know, connect and kick around ideas because we're at probably the most fun time in my career, at least I've been Silicon Valley since 2000, where we truly are inventing the future. And, you know, so if you jump in, now is your chance to like, shape what the next years are going to look like. And I think very few points in history kind of come along like that. So now's one of those times when it's just really fun to kick around ideas and find something to tackle and just build it and make it real.
- [26:27]
Nathan Isaacs
Awesome. So thank you so much. I appreciate it.
- [26:32]
Matt Howland
Awesome. Thank you.
- [26:34]
Podcast Host / UserTesting Narrator
Want to keep the conversation going? You can find the show notes@usertesting.com podcast if you haven't already, don't forget to follow us on Apple Podcasts, Spotify, Overcast or Google Play so you never miss an episode. And if you enjoyed today's show, please share it with a friend or leave us a rating and review on Apple Podcasts and on until next time, this is Insights Unlocked, an original podcast from User Testing.