Skip to content
PodSubs
ShowsPeopleGuestsJobsNewsEvents
More
BlogPodCred
Menu
ShowsPeopleGuestsJobsNewsEventsBlogPodCredNetworksSign in
Join nowSign in
  • Shows
  • People
  • Jobs
  • News
  • More
    • Guests
    • Events
    • Blog
    • PodCred
    • Networks
    Join nowSign in

Retailgentic | Consumer Behavior & Retail Trend

Retailgentic | Consumer Behavior & Retail Trend artwork

AI Is a New Customer Category: Wharton Professor & Bodhium Labs Co-Founder Kartik Hosanagar

Jul 23, 202643 minAudio
ShareXLinkedInFacebookWhatsAppEmail

Kartik Hosanagar wears a lot of hats: Wharton professor, serial entrepreneur, angel investor, and now co-founder of an applied AI lab building simulation tools for AI search. Scot and Kartik land in almost exactly the same place on where agentic commerce is heading, but they get there from opposite directions: Scot on gut and operator instinct, Kartik on research. The result is a wide-ranging conversation about treating AI as a customer, the levels of agentic commerce, and what brands should actually be doing this week, this quarter, and this year. Highlights: AI is a new customer category: Once shopping goes agentic, Kartik argues you have to ask of AI what you ask of any customer: how do we inform it, persuade it, and what biases does it have? The driverless-car model: Agentic commerce has levels 0-5, just like autonomous vehicles. Level 5 is “buy the shoe and make sure it arrives in a week” hands-off; Instacart-style automation and AI-enabled discovery are lower levels and we’re already in the era. The funnel got outsourced: Top-of-funnel work customers used to do themselves like discovering brands and building a consideration set is now done by AI, so shoppers arrive far later in the funnel. Discovery is “fundamentally transformed.” 4 words → 60 words: From his CommerceNext keynote, average search went from about four words on Google to about 60 words on the first ChatGPT turn, then multi-turn. Nobody types “men’s running shoes” anymore; shoes show up on conversation turn seven. UCP, ACP and the protocol wars: Kartik explains Google’s UCP (product feeds today, transactions later), why OpenAI backed off ACP instant checkout, and reads Amazon joining UCP as “testing the waters,” not a strategy reversal Counterfactual simulations: Bodhium builds “digital twins” of ChatGPT and Gemini and spins up virtual worlds where your site looks different, testing a thousand hypotheses in minutes to find what makes AI answers more accurate and favorable. Market research, but for AI. The today / this quarter / this year roadmap: Now: audit conversations, visibility, and which bots read your site. This quarter: build simulators and a UCP/MCP workstream. This year: scale it with agentic workflows. Two front doors: Kartik doesn’t think websites disappear, but “a website is not the right front door for an agent.” Expect a human web door and an MCP door for agents, plus omnichannel behavior where each feeds the other. Kartik calls himself an "agentic commerce maximalist," and after this conversation you'll see why... the research and the gut instinct point the same direction. Enjoy the conversation. Timestamps: 03:04 Welcome: the "renaissance man" background problem 03:49 Kartik's cocktail-party version: academic + entrepreneur, Yodle 04:36 The Yodle / Milo mobile-social-local era (and Fidji at OpenAI) 06:05 Wharton AI Center and what Kartik researches 07:37 The thesis: AI search is more than another channel 10:08 Defining agentic commerce: the driverless-car levels 12:34 UCP, Gemini, and why Amazon joined 15:55 Amazon vs. Walmart strategies (Rufus vs. Sparky) 16:29 Bodhium Labs: what it is 17:00 The AEO/GEO problem: guesswork and the Reddit trap 18:56 Research-led counterfactual simulations 21:02 The long tail: multi-turn conversations replace keywords 22:49 CommerceNext keynote: 4 words to 60 words 24:57 The "AI is just another channel" misconception 26:48 Shopper trust: high for discovery, not for checkout 28:15 You can't fully control what AI thinks about you 30:51 The roadmap: today, this quarter, this year 34:41 Rufus, Sparky, UCP, and MCP endpoints 36:37 Does agentic commerce grow the pie or shift it? 38:32 Do we still need websites? Two front doors 40:39 Where to find Kartik: LinkedIn and Substack 👉 Connect with Kartik: https://www.linkedin.com/in/kartik-hosanagar-22272115/ 👉 Learn more about A Human’s Guide to Machine Intelligence: https://www.amazon.com/dp/0525560882?lv=shuf&channelId=500&plpRedirect=mhFallback 👉 Visit his substack: https://hosanagar.substack.com/ 🧠 Want to stay ahead in AI commerce? Subscribe and follow along: 📰 Subscribe to the free Substack: retailgentic.ai 📺 Watch episodes on YouTube & Subscribe for updates: https://www.youtube.com/@Retailgentic

Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.

Transcript

From the show. Tap a time to play from there.

Speaker 1For a lot of movie studios, Netflix was another channel through which you could distribute your movies. But it was more than another channel because what some of these technologies can do is they can change entire systems of work. And so what happened was that Netflix became basically so powerful that they restructured the entire content industry and determined what gets made and who gets to watch it. AI enabled commerce or agentic commerce is a bit of that, and so you've got to really think about how's the value being restructured, who has more power as a result of that, and things like that.

Speaker 2Welcome to the Retailgentic podcast where we explore the intersection of retail, ecommerce, and AI agents which we call Retailgentic. Our mission is simple. We are shining a bright light on the new Agentic AI shopping megatrend which will disrupt the trillion dollar digital retail market. The stakes are high. Google, OpenAI, Microsoft, Perplexity, Amazon, Apple, Meta, and Anthropic are all fighting to be the consumer AI shopping agent winner.

There are gonna be some tough decisions for online brands and retailers to make. What happens to checkout? How should payments work? Should you continue to invest in retail media networks? What does the world look like when over half of your traffic is not from humans but from AI agents? These existential questions are the tip of the iceberg. On this podcast, we're talking to industry experts, analysts, and influencers to get answers to these questions, identify trends, and best practices. This is Retailgentic, your podcast partner in navigating our AI shopping agent future. And now here's your host, Scot Wingo.

Speaker 3Welcome to this episode of Retailgentic. My guest today is Kartik Hosanagar. Kartik is a Wharton professor, serial entrepreneur, angel investor, renaissance, a Gentic comerser, and he also runs the Wharton AI Center. In addition to all that, he is also the founder of a company that does AI simulations around AEO GEO called Bodhium. Some of our most popular podcasts here on Retailgentic feature academics like Kartik that have done a lot of research in the area, And I think you're really gonna enjoy my conversation here.

It turns out we, Kartik, I share a lot of the same beliefs and conclusions of where agentic commerce is going, but we come at them from very different angles. And, I have more of a gut feeling, and he has a lot of research to back up what he's doing. And it's always really fun, and fascinating to explore everyone's different path to their beliefs on agenda commerce. He spends a lot of time talking to companies about it as well. He recently gave a keynote over at Commerce Next with our friends there.

And I don't want to spoil any more of the conversation, so I'm going to pause here and let you enjoy my conversation with Kartik. Welcome to this episode of Retailgentic. My guest today is Kartik Hosanagar. He is a serial founder, professor at Wharton, angel investor, and a PhD from Carnegie Mellon. Kartik, welcome

Speaker 1to the show. Scot, thanks for having me.

Speaker 3Yeah. You're a renaissance man. I was researching, I was like, wow. The background section of this is gonna take an hour on its own. You've you've been very busy.

Speaker 1Or unfocused, depending on how you look at it.

Speaker 3No. There is a theme. There's a common theme in there. And I'm surprised we haven't met. We've come close a couple of times in there, which we'll hit on. But in your own... That was like some of the highlights. When you're at a cocktail party, how do you tell folks about your background and what you're up to now?

Speaker 1Yeah. Look, I've been an academic and an entrepreneur for the last twenty five years roughly. So as an academic, I'm a professor at the Wharton School. I run our AI center as well there. And as an entrepreneur, was previously a co founder of a company called Yodle, which was AI applied to online marketing. More recently, I'm a co founder of Bodhium Labs, which is an applied AI startup, which is also focused on using AI to improve marketing. I'm sure we'll talk about that in a bit. Other than that, I've been an author of a book called A Human's Guide to Machine Intelligence and an angel investor in a a few dozen start ups.

Speaker 3It's funny. My daughter, wanted to watch Silicon Valley, so I haven't seen that in a very long time. So I was watching it, and, as I was preparing for the show, they had the episode where they're at tech tech crunch disrupt, and every pitch is so low mo. Do you remember that? And it made me think of that Yodle Milo era when you guys were in the mix of... Yeah. You kinda had to be mobile, social, and local, and you guys were always... You you you had two companies that played in there.

Speaker 1That's right. That's right. Yeah. Yodhl was a a local search, and then, Milo, where I was an adviser, I was a local, and mobile, kind of player. So, yes, I guess, you know, I haven't seen Silicon Valley in a while, but they have good parodies on what's going on.

Speaker 3It still it still really lands. It's pretty good. You know, it's pretty funny. The, in the AI era, a lot of it's funny. Like, there's one where Guilfoyle's AI, like, orders... Goes crazy and orders like a... It's supposed to save money on hamburgers, but it orders like a pallet of hamburger meat. It's kind of funny they were very predictive about a lot of things in our world today.

Speaker 1A lot of things indeed. I remember the one episode where this guy's in a driverless car, but the driverless car won't let him out and takes him into a ship. A container. Yeah.

Speaker 3Milo was interesting. So I had ChannelAdvisor, and we worked with them to get the local inventory out, and then publish it, before they were acquired by eBay. And then Fiji ran that when she was at eBay. So it's kind of funny. She... She's now like the head of all the consumer stuff at at OpenAI, as you probably know.

Speaker 1Yeah. Very cool.

Speaker 3So tell us... Let's start at the academic side. What are some interesting things that you're thinking about at Wharton, and what does the AI center do?

Speaker 1You know, at the school, really, there's a lot of things AI related that we think about, everything from what do our students need to know in an AI era where it's unclear what will be the skills of the future, so how do we redesign our curriculum around AI? What are the AI skills we need to impart to our students? Our center itself is a bit more research focused, so we support faculty research on these topics. Personally, I'm focused on a few different themes. One area is around designing human AI collaboration.

Another area is on AI search and how people are moving from typing three, four keywords on Google to having really long essays and conversations with ChatGPT and Gemini and how that's changing search in general, but, in particular from a business perspective, how does it change shopping in particular and how people discover products and brands and things like that. So, yeah, these are a couple of themes that I'm personally focused on.

Speaker 3And that is why you're on the podcast. That's a... This is a podcast that's exclusively about that. So what are, you know, what are your thoughts? You had a post kind of back in January that was pretty early on with some pretty interesting thinking around. I think in that post, the the the crux of it was talking about the the retailer hosted answer engines. So back then, it was Rufus and Sparky. Now it's Alexa for shopping, which is much more of a mouthful. But walk through your thesis there and then maybe update it for modern times, which is... I know it's only July, but things change so fast.

Speaker 1Yeah. Look, I think there's there's many layers to peel there, but at a high level, I would say my thesis is that any retailer or brand that's thinking about AI search as merely another channel that customers are using to discover the brands are missing the big story because these systems are more than another channel through which they discover your brand. But increasingly, people are outsourcing a lot of the thinking to these AI engines. For example, previously, you'd go type a Google query and get 10, 20 results, click on a bunch of them, discover a bunch of brands, synthesize that information, make sense of it all. You might buy in one place, but you did all the top of the funnel work yourself.

You discovered a bunch of brands and so on. Even if you bought from one, you discovered many others. Now increasingly that is being outsourced to AI, so what a customer used to do on their own is being done by AI and the consideration set formation is done by AI. And so people come in much later in the funnel. So discoveries fundamentally transform. And as these things become more agentic, and we know the AI labs are spending a lot of money on trying to create agents that can do shopping for people, once it becomes more agentic, we should start thinking about AI as almost as a new customer category itself. And so you have to ask, just like with your customers, you ask, how do we inform these customers? How do I persuade these customers? We think about what influences these customers. You need to ask the same question about agents, which is things like how do we persuade AI, what drives AI decision making, what biases do AI have?

We understand $9.99 is different than $10 for humans. Is that true for AI or not? We don't know. So these kinds of things. So we need a new thinking and a new discipline to think about, like, AI decision making as it relates to commerce.

Speaker 3Yeah. One of the debates we commonly have on this show is the definition of agenda commerce. So there's there's a set of people that that they have a very strict definition that agenda commerce has to be totally hands off the keyboard. You know, Kartik wakes up, says, I need a new shirt, and the machine just goes and does it without any input. I'm in the camp of I kind of have a bigger tent where if you're using it anywhere along the journey, I kind of count it. But where do you fall on that spectrum of the definition of Attemptive Commerce?

Speaker 1I'm fully with you. For me, the analogy I would use is that of driverless cars. So with driverless cars, we talk about level zero, level one, level two, level three, level four, level five. And level five is where you don't even need a steering wheel in the car. You can sit at the back and the car takes you wherever. And that's the goal of many companies, but that doesn't mean if you're sitting in a Tesla, you're not in a driverless car or rather in an autonomous vehicle. So I think there are levels to it. So yes, you can have a true agentic level five commerce where you kind of just say into your phone, hey Siri, I'm looking to buy a shoe under $100 for my half marathon coming up next weekend, figure it out, buy it, order it, and make sure it arrives in a week. And then it does the rest. But that doesn't mean that you cannot have intermediate things.

And in fact, the path to level five agentic is going to involve these steps along the way. I feel level one is even things like Instacart where you can have a lot of automation around your ordering. And level two for me is AI enabled discovery, where you can go in and ask questions to ChatGPT and it'll give you the shortlist and it'll explain, here are the four brands of shoes I recommend, here's why, and based on all these factors, this is the one that's right for you. And then you can, of course, go further down the funnel and ask, can you do AI checkout? And does a human need to be involved for approval?

And so on, and you can build that out. But I'm with you that just because something is not end to end AI enabled doesn't mean that we haven't begun the agent e commerce era.

Speaker 3And then what do you think about where the answer engines are today? So we've got, Gemini has... There's kind of a UCP group forming, and there's there's actually, like, a consortium. It's probably not the right word, but a council and stuff. Number one, a lot of people were surprised Amazon joined that, so no one really knows exactly what they're doing. Are they gonna are they gonna publish out to UCP, or are they just gonna ingest it? So so we'd love to hear your thoughts on UCP, and then then we'll go to the next one. What what do

Speaker 1you think about UCP and Gemini and and all that? So UCP or Universal Commerce Protocol is basically Google's big initiative so that agents can talk to each other and do all the things that are involved in commerce. One of those basic things is even exchange product feed and product information. So for example, a merchant that decides, hey, I'm going to make my catalog available in a UCP compliant manner, will make it super easy for any other agent to use UCP to read a product catalog and pull it in. So for example, Shopify, among others, are kind of saying, hey, Shopify merchants will make them UCP enabled so anyone can pull their product catalog information in and make it available on any AI engine that decides to follow UCP.

Of course, Gemini would be one of them, could be anyone else. So it's an open standard. Somebody else could also build an AI discovery platform and use UCP even though they're not Google, they can use UCP to pull the product feeds. So there's that piece. Then there's the whole, can you transact with it? Can you pay with it? And so on. And UCP gets into those pieces as well. And those are where we get speculative about, like, hey, will people want to transact with AI without any human intervention? And we've seen OpenAI launched agent e commerce protocol, or ACP, and it's still around mostly from a product feed exchange perspective.

But they've backed off on the instant checkout where they were saying anyone can check out on OpenAI and OpenAI's agent will talk to the merchant's agent and finish the transaction. And that hasn't played out. I think it's going to take some time before agentic checkout and agentic payments are a reality. But that's kind of the vision of Google. And coming to your point about Amazon, it's unclear what their strategy is. I think one the one thing that's pretty clear is they don't want to be in a world where agentic transactions are happening on other platforms like OpenAI and Google because then OpenAI and Google start to have disproportionate influence on purchases happening within Amazon. And so historically, they have resisted crawlers from OpenAI and Google from coming and scraping their site and pulling product information. And they've launched Rufus and their entire bet has been Rufus, which is their agent for commerce and so on. And their strategy has been very unlike Walmart, and Walmart has said, oh yeah, we'll do ACP, we'll do UCP, we'll have our own agent, Sparky, but we'll go ahead and use, make it available on those other channels as well. So now for Amazon to say we'll be on UCP, I have a feeling they just wanna know what's going on there.

They're kind of testing the waters. I don't think it's a reversal of their strategy. They're testing the waters, and they're trying to make sure they're informed about what's happening in that camp. But their strategy to me seems to be we want any agentic transactions happening through our own AI agents. So anyone who wants to do agentic commerce, come on Rufus and do it there. And we own things end to end. I believe that's their strategy.

Speaker 3You have your academic side and then your professional side, and then you have Bodhium. Tell us a little

Speaker 1bit about that. Bodhium, as I mentioned, is an applied AI lab. And we are trying to solve a few different questions around AI enabled commerce. And the first one that we've tackled is a market that is sometimes referred to as AI engine optimization, or AEO, or generative engine optimization, or GE. And what we found in that space is that brands are interested in understanding what they can do to make sure that they can inform AI engines about their product offerings well and also persuade those AI engines that they have the right product for the right customer.

Today, what's happening in that space is it's all guesswork. There's dozens of companies in this space and all of them do the same thing. They will take your brand and they will come up with 200, 300, 400 questions that your customers may potentially be asking about your product category on OpenAI or on Claude or Gemini. And they will tell you whether you show up in the answers to those questions and whether your competitors show up in the answers. So for example, they will inform Nike, hey, if somebody says I'm running a half marathon and I need a shoe that is good and has strong ankle support, do you show up?

Or is it your competitors? I'm wearing a HOKA right now, so does HOKA instead dominate that? And things like that. But what are you supposed to do if you feel like you deserve, you have a right to win that answer, but you don't show up? And nobody knows the answer to that question. Everyone says, oh, why don't you create more content? Why don't you create more blog posts and see what happens? That's like throwing spaghetti on the wall and seeing what sticks, and it doesn't quite work.

Speaker 3Or my favorite is go start a discussion about your brand on Reddit. Yeah. That's kind of crazy talk.

Speaker 1I hate that. Yeah, exactly.

Speaker 3Yeah. Everyone always pushes the Reddit, yeah.

Speaker 1Pretend you're a real Reddit user and start hyping up your brand and all

Speaker 3of that. Yeah, some of it gets really sketchy pretty fast. Very gray hat to be generous. Yeah.

Speaker 1We kind of said, look, I think there's an opportunity to build something that is research led. My co founder was previously at DeepMind for the last six and a half years, has been in search for the last twenty plus years. And we kind of said, Look, I think the interesting thing is that these AI models can be probed, meaning you can interpret their answers. You can ask, why is the answer what it is. And if you can interpret them, then you can start to ask, hey, what might happen if I did this?

Or what might happen if I did that? And so we're building a platform where you can run what we call counterfactual simulations, where basically we create virtual worlds where in each virtual world your website looks different. And so our AI agents create a new virtual world where there's a new website, a new look and feel. And we test various hypotheses in the virtual world where we'll have a digital twin of, say, ChatGPT, and we'll ask, what might the answer on ChatGPT be if your content look not like A but like B? And through a series of these tests that can run-in a matter of minutes or hours, we can discover, having tried thousand different hypotheses, we might discover one or two where the AI answer is more accurate, more favorable to your brand.

And in that, we kind of identify how you present your information on your website so it influences AI. It's not unlike market research where you might test with a bunch of customers and you'd ask them, hey, which one do you like better, this messaging or that messaging? But we test that with AI and we kind of give you the information on what's going to work or not. And that's our first product. And over time, we'll expand it to other things broadly in the space.

At the end of the day, if you can understand AI answers and have interpretability and explanations around AI answers, then you can do a lot of things around it.

Speaker 3Kind of back to the front part, another thing I think the GEO companies do a disservice is they they kinda want it to look like SEO, and, you know, they want it to look like the keyword era. So they they they make it seem like they're still ahead and and tail distribution curve. And I'm I'm pretty convinced that doesn't exist anymore because I think every customer journey is essentially long tail. People aren't using these these answer engines to just kinda say men's running shoes anymore. Some of them may be, but once they get... Once the user gets pretty sophisticated, they're they're having a multi turn conversation. And many of the conversations, if you look at the OpenAI papers, the, you know, the conversation is I'm taking a trip to somewhere, and then somewhere in there, it says, oh, if you... If you're gonna go hiking, you need to think about this kind of kit. And then, you know, and then conversation turn number seven is shoes. It's very different than the keyword era.

Do you guys have any evidence that supports that? It seems like you're simulating merchant back, not consumer down.

Speaker 1Yeah, we're simulating merchants, but in the process of that, we have to simulate customer behavior as well. So we have to simulate customer personas who have a job to be done, who will have a multi turn conversation with LLMs and so on. The one thing we can say is that you're spot on, the conversations are not men's running shoes. So in fact, I recently presented my keynote at the Commerce Next conference where I was talking about the fact that the search queries have gone from four words on average on Google to 60 words the first turn or the first chat on ChatGPD. And then you have, you get the answer, then you ask another question and another follow-up and so on.

And what's happening is exactly what you said, which is people are moving from high level top of funnel to late funnel completely on the AI platform. And the consideration set is formed on the AI platform through a multi turn conversation. And the conversation, well, I'll stick to the running shoe analogy. The conversation starts with, excuse me, I am running a half marathon in a few weeks. What should I do to prepare for it?

And in there will be suggestions like your training regimen in preparation for it, suggestions around your diet in the days leading up to the half marathon, suggestions around electrolytes, suggestions around running shoes, and so on. And then different people will have different subsequent questions. Someone in there would say, oh, tell me more about what do I need in a running shoe. And then they might say, oh, I do not currently, or they might say, my current running shoe is whatever, it's a Nike, it's one of the Jordans, what do you think, can I run with that? And then they might say, no, that's not good for a half marathon.

Then you might say, oh, what do I need in a running shoe for the half marathon, and which is the best brand? So as you were saying, it's much later in the conversation. And so really for brands, it's a completely new world where you have to think about how do I educate AI about what products I have, where I'm relevant, how to insert my brand at the right point in the conversation with the customer. And so that requires much more than, here's my PDP and here's an FAQ on the product itself.

Speaker 3What else? Any other interesting takeaways from the Commerce Next talk that you did since it's probably fresh? That was, like, last week, I think, so it's probably fresh in your mind.

Speaker 1Yeah. Yeah. I think a few things came up in there. We had some conversations around, like, what are some mis conceptions that retailers and brands often have when it comes to AI. And a few came up. One was something we talked about, which is AI is another channel. And I talked a little bit about why that is potentially harmful. And it's correct, it's an AI channel, but it's dangerous because it's incomplete. And the analogy I gave them was, for a lot of movie studios, Netflix was another channel through which you could distribute your movies. But it was more than another channel because what some of these technologies can do is they can change entire systems of work.

And so what happened was that Netflix became basically so powerful that they restructured the entire content industry and determined what gets made and who gets to watch it. And so I was kind of highlighting that AI enabled commerce or agentic commerce is a bit of that. And so you've got to really think about how is the value being restructured, who has more power as a result of that, and things like that. The other thing that came up is also this idea that you have to start thinking about AI as a customer as well. And so this idea that there will always be a human at the other end, you've got to question that as we were talking about this movement towards this level five automation.

And then ask, okay, if AI is the customer, what do I know about influencing AI? That's the other piece that came up.

Speaker 3Have you done any research on trust, shopper trust with AI? Is that an area of interest to you?

Speaker 1Yeah, absolutely. Yeah, I think when it comes to shopper trust with AI, the evidence is that when it comes to discovery, there is a lot of trust today. I think people, you ask them about, do you know AI can hallucinate? They are aware, but the reduction in search costs, the time saving is so significant that they're willing to accept the occasional hallucination and do a lot of their research there. Trust when it comes to full automation and checking out on AI systems is not quite there.

People still want to click through, visit the merchant website. And so when it comes to the full agentic commerce vision, I think there is a big trust gap today. And people are not yet ready to do anything late funnel, certainly not outsource the final decision to AI, certainly not do it in a no human mode today. And also when it comes to payments and things like that, there is some reluctance. But otherwise, all the trends are discovery, trust is very high.

Speaker 3Any other areas of interest there for you?

Speaker 1Yeah, I think one of the things that we're also seeing is that AI can certainly presents a new interesting opportunity for brands where if you're early and you play the AI visibility game well, then you could show up and be a prominent brand in AI search results. But we also see some interesting challenges where you cannot fully control what AI thinks about you. So two examples, we had an e commerce retailer from London who was asking us, help us understand what AI thinks about our brand and so on. And we kind of told them after an analysis, AI thinks you're actually, you've shut down. And then they were surprised and we talked about it and we were like, well, here's what AI is saying, your brand shut down.

And then they kind of said, oh, the history is that the previous owner of the brand was struggling financially and they decided to put the brand up for sale. And we bought the brand. But then there is some discussion online that their stores might be shutting down because they were financially distressed. And these conversations in the community that the stores might be shutting down was sufficient for AI to believe that it's shutting down. And so while ChatGPT thought highly of the brand, it would not recommend the brand because it was thinking it was going to shut down.

Another example was a wallet company that became a top recommendation for best gifts for men, top wallet recommendation for searches around things like gifts for men and so on. But AI was also giving wrong information about promotions they have and so on. And that information was coming from discussions online. And so we were talking about Reddit earlier. It is true that you cannot fully control your brand narrative and these long form answers are determined partly by what you release on your website, partly by what third party media property release on their websites, partly by what the community discusses.

And so I think we're in an interesting era where companies have to think more holistically about all of the three. But it's not sufficient to say, We'll create some Reddit profiles and post about our brand. It's got to be a much more responsible engagement with the community.

Speaker 3What should brands be doing today? What's some of the actionable things that they can do to make sure the AI is picking them and that it's not making these kind of crazy recommendations?

Speaker 1Yeah. So in the keynote, I did... We spent some good period of time talking about what a twelve month roadmap for AI could look like. And one of the things I was kind of emphasizing in that keynote was that we need to think about what are you already doing today, and if you aren't, you should be doing it this week. What are the things that we should be doing this quarter, and then what are the things we should be doing, say, this year?

And so I'll break it up into these three buckets. So in terms of what are the things you should be doing right now, at a very basic level, you should be fully understanding what are the types of conversations your customers are having on these LLMs. Secondly, who shows up, which is the AI visibility analysis. Do we show up? Do our competitors show up? And where do we show up? What are our strengths and weaknesses? We do well with wallets, but we don't do well in, let's say, also sell hats and tech gadgets, we don't do well in those categories. So understanding our areas of strengths and weaknesses. Which bots are visiting our website?

How frequently do they visit our website? What are they actually reading on our website, so what are the bots looking for? And by bots, I'm specifically interested in the AI engines like ChatGPT, Gemini, Claude. And every time a user searches something on the site, they will have agents called user agents visiting retail properties and brands' websites to look for information. So what are they actually scouring? What are they trying to read? So that should be done right now. I think this quarter I would ask every company to be really thinking about interpretability and explanations on, now that you've done an audit, why do we not show up? What will happen if we make certain kinds of changes? Every company is being told several suggestions like, hey, you need to have an FAQ.

You need to have information presented in the form of tables. Your blog format should be a headline question and an answer and then the rest of the blog. And lots of things, right? Your schema should look different. Your PDP should look different and so on. So there's a lot of hypotheses. So within this quarter, you should go from these random guesswork type of hypothesis, he said, she said type of hypothesis, to creating simulation environments where you're able to test these strategies. So you should have your own simulator of ChatGPT, your own simulator of Gemini, and you should be able to test these hypotheses. And by the end of the quarter, you should know which ones are working for you, which ones are not working for you. And that's going to be important.

And so you have to do that this quarter. And by, I think, the next twelve months is going to be about scaling what I'm just describing. So how do you bring in agentic workflows where agents can analyze your visibility data, agents can come up with hypotheses of changes, agents can use the LLM simulators to test the hypothesis and come to you with recommendations on make these changes, here's why, and here's what I expect will happen. And that's how you can scale this. And if any company can pull this off in a year, I feel you'll be miles ahead and you'll be also improving your AI visibility and your AI game so rapidly relative to your competitors who will still be playing the let's throw spaghetti on the wall and hope something sticks kind of game.

Speaker 3So that's the answer, engines. Does this kind of count for the retailer kind of hosted engines as well for Rufus and Sparky? Should you do the same kind of thing?

Speaker 1Yeah, sorry. And I think on the Rufus and Sparky side, I think my version of that would be today, as I said, look at which bots are visiting your site, what are they doing. This quarter, have some sort of a team in your company that is looking into UCP and looking into how do we make sure our product feed, for example, is UCP compatible so we can submit our product feed to various engines on UCP. Or if you're on Shopify, how do we leverage Shopify's UCP capabilities to make sure our product feeds are fed to the engines that are looking for our product information? I would also look at an MCP endpoint this quarter.

So MCP is essentially another protocol for agents to go and fetch information. So an MCP endpoint would be a way for agents to come, instead of visiting your pages, coming to your MCP endpoint and understanding your product information, your brand information, and getting it there. So those are the kinds of things I would investigate. I think it's too soon to say that there'll be agentic transactions, but I think at a minimum you need to do that. And if you are selling on Amazon, you're selling on Walmart, you need to have a roofer's work stream as well.

So understanding how we show up for popular queries on roofers, what are the things we can do to influence roofers. So it's a very similar what can we do today, what can we do this quarter, what do we do this year, But translating that to roof first, UCP, and MCP would be how I think about that in the agentic sense.

Speaker 3How about let's... So that's kind of the tactical view. Let's let's look forward a little bit. You know, there's a lot of forecasts out there that say by 2030, Agentic Commerce will be somewhere between 1525% of e commerce. I also... I get excited more by the... That's just kind of you know, swapping swapping, you know, things across buckets. There are some forecasts that show this accelerating e commerce. Do you think agentic commerce accelerates e commerce in its adoption rate? And then where where do you think this is all going?

Speaker 1I think of agentic commerce as essentially reducing a lot of friction for customers and for shopping. And so once you reduce frictions, you would expect some level of increase in buying activity, meaning people actually either have an interest in buying something but they can't find it, or it's just too time consuming to go discover that, they're not aware of something that they could actually use. So yes, there's going to be some level of increase. But to me, the analogy is similar to e commerce reducing frictions. And the primary impact of e commerce was taking away share from physical commerce towards e commerce.

And yes, there is some increase as well, so buy did grow a lot. But the more significant impact was that of taking share away from traditional commerce. I would view agentic commerce the same way. Yes, there will be some increase in the buy because it made things more frictionless. But the primary impact and the bigger impact will be that of what share of commerce was taken away that might have happened in more traditional digital ways like search and discovery in the usual manner versus agentic ways.

But I think the agentic lower friction leads to more purchases is a real possibility as well.

Speaker 3There's a set of people that kind of are in this camp where, well, maybe you don't really need websites at the end of the day. If these agents are, like, just crawling your website, wouldn't it be more efficient to just have agent to agent communications? Do you do you see that as an endpoint here, or will humans always wanna get to a website and poke around on their own?

Speaker 1Yeah. I I think it's going to be a little bit like, well, stores haven't gone away. Stores are still there, and websites are still there. And in fact, you look at companies like Wharton that started as pure e commerce companies, they have doubled down on their store strategy. What they've realized is, you know, there are some people who only shop in stores, some people who only shop online. And by the way, there are omnichannel shoppers. And actually, opening stores increases online sales, and online website increases store sales. Omnichannel shoppers like both. I think it's going to be like that. Let's say agent e commerce takes off by 2030, as many are projecting and predicting, which is definitely in the realm of possibilities.

It really takes off. I don't think the website goes away. There's going to be some people who are going to use websites and shop on their own. There are going to be some people who will use agents to do the work, and they will need a separate front door. I don't think a website is the right front door for an agent. And that's why I was talking about MCP endpoints, where you have a separate front door for agents that can come in and don't have to transact and use a browser the way we do. And you might have to have two separate front doors. And you might discover there are some omnichannel people, the equivalent of omnichannel people, where sometimes the website helps feed the agentic store and sometimes the agentic store feeds the website because there is going to be some human AI collaboration as well where people will ask agents to do certain pieces but then come in for certain other pieces and vice versa. So I think there's going to be room for both front doors. I don't see the regular web front door disappearing.

Speaker 3Cool. If people want to learn more about your research and kind of your thoughts where, obviously, you're pretty easy to find on LinkedIn and you're pretty active there, where else should they look?

Speaker 1Yeah. I think I would say, first and foremost, LinkedIn. I'm, as you said, quite active there, and I share my opinions and thoughts on this space and on AI and business in general over there. I also run a newsletterblog, which is on Substack. So you can go to my last name, Hosanagar, hosanagar,.substack.com. And that is the link to my Substack, and that's where I also have long form posts on these. So sometimes I will do quick posts on LinkedIn, but sometimes I'll feel like, hey, this one deserves a much longer one. So once a month, I'll do a longer post on Substack. And so that's the other place where they can find out, about my work and my thoughts on on these topics.

Speaker 3Yeah. We'll put a link to that in the show notes as well as your LinkedIn. And yeah, I'm a avid reader, and that's how I knew to reach out to you. So I really appreciate you putting your thoughts out there.

Speaker 1Thank you very much. I appreciate, anyone who's taking the time and the trouble to read these longer posts. I appreciate that.

Speaker 3I'm a... I've been called an agentic commerce maximal... Maximist, so it is part of my job description now to read everything that that people put out.

Speaker 1That's a great choice you made because I'm very bullish on on these trends. And like we discussed, even if level level five automation doesn't happen, there's no doubt in my mind there's a lot of agentic automation in commerce happening, and we should all, you know, increase our understanding of this space.

Speaker 3Thanks for coming on, and good luck in all of your many endeavors.

Speaker 1Scot, thanks so much for having me.

Speaker 2Follow, rate, and share to stay up to date with Retailgentic, your guide to the future of AI shopping agents.

Open the transcript file

Hosts and Guests

Hosts

  • Scot Wingo Host Feed-stated

Credit the people on this episode

Be the first to rate this episode

More episodes

All episodes
  1. "Every Ad Will Be an Agent," InMobi & Glance Founder & CEO Naveen Tewari

    Nov 16, 202647 minAudio

    This week we go international. Scot is joined live from India by the founder of one of the largest agentic shopping platforms you may not have heard of yet. This one is part founder story and part product deep dive: how...

    Credit the people on this episode

  2. How Brands are Leveraging AI in the Design Phase: Raspberry AI Founder & CEO Cheryl Liu

    Nov 12, 202645 minAudio

    About every six months on Retailgentic, we step away from the front end of agentic commerce to look at how AI is changing a different part of the retail stack. This week it’s product creation. Raspberry AI helps fashion...

    Credit the people on this episode

  3. AI Meets Fashion: Co-Founder & Chief Brands Officer Lisa Yamner on Daydream’s Big Bet

    Nov 5, 202648 minAudio

    When you think about vertical shopping agents, the AI agents built to solve shopping in one category, one name keeps coming up: Daydream. We’d wanted to get the Daydream team on Retailgentic for a while, and this week...

    Credit the people on this episode

  4. Shoppers Take the Agent’s Top Pick: Growth Memo Author Kevin Indig on How People Really Shop in ChatGPT and AI Mode

    Oct 29, 202645 minAudio

    This week is a research-heavy episode with a conversation of one of the sharpest observers of how real people search and shop now that AI sits between them and the products they want. Kevin’s user-behavior research...

    Credit the people on this episode

  5. The Missing Link Is Shopper Context: eComID CEO Oscar Rundqvist; Plus eComID’s Big Funding Announcement 📣

    Oct 22, 202644 minAudio

    This week is a product deep dive with a fresh funding story attached. eComID just closed Europe’s largest fashion tech seed round, and the company is building something agentic commerce needs: shopper context that...

    Credit the people on this episode