In this episode of Retailgenic, Scot Wingo welcomes Ananda Chakravarty, VP of Research at IDC’s Retail Insights Group. Ananda shares his decades of retail and technology experience spanning Monster, Staples, Oracle, Diebold Nixdorf, and Forrester to help break down: The real definition of Agentic AI How retailers must prepare their data and sites for AI agents Why product discovery is ripe for AI-driven transformation The massive implications for pricing, loyalty, personalization, and retail media How agent-to-agent commerce may soon reshape buying behavior entirely This is a must-listen for anyone thinking about the future of retail, commerce, and AI. 📰 Subscribe to the free Substack: retailgentic.ai 📺 Watch episodes on YouTube: youtube.com/@retailgentic 🎧 Listen wherever you get your podcasts: Spotify, Apple, etc. Highlights: 00:02:15 Ananda explains his IDC report on agentic AI & product discovery 00:05:00 His career journey: engineering to business to retail tech 00:10:45 Early e-commerce work at Monster, Staples, Talbots 00:26:00 Deep dive into agentic AI orchestration & multi-agent workflows 00:36:00 What retailers should do today to prep for AI agents 00:41:00 How pricing, loyalty, and personalization will change 00:44:00 Agents negotiating with agents (agent-to-agent commerce) 00:48:00 The rapid acceleration of agentic traffic vs human traffic 00:49:00 Where to follow Ananda's work
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.
And Microsoft's Copilot is really an interesting play, uh, because it's now no longer, uh, that coded environment that you have for, for developers anymore. It's just, "Oh, tell me what I want," and, and there you go. You got this program and code that can do exactly what you want. I wanna go scrape 1,000, uh, websites for price points to compare it to what my price point is. Here you go, uh, and, and just by asking the question. And that's the kind of thing that I think is gonna start changing the work environments for, uh, retail, uh, employees as well
as other businesses as well. But m- mostly you're gonna see a lot of the impact on the retail side Welcome to the Retailgentic podcast, where we explore the intersection of retail, e-commerce, and AI agents, which we call retailgenic. Our mission is simple. We are shining a bright light on the new agentic AI shopping mega trend, 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 checkouts? 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. Welcome to the Retailgentic podcast. IDC is a leading global provider of market intelligence and advisory services. They work with a wide range of businesses from B2B to consumer and across all different sectors. In today's episode, I'm excited to welcome Andy Chaurtia. He is the vice president of research for IDC's Retail Insights group.
Andy has a report out titled "Win-Win Product Discovery: A Retail Agentic AI Use Case." In this report, Andy goes very deep into what agents are, how do you orchestrate them, and how retailers can leverage agents internally to improve their product catalog discoverability. I thought it was a great example. He uses baseball bats, which is, uh, a category that is inherently hard to have a great shopping experience for in today's filtered navigation user interface that most retailers have. After reading this awesome report, I thought, "We have got to get Andy
onto the Retailgentic pod," and he graciously accepted that invite. In this discussion, we hear all about Andy's long and amazing career, where he was on both sides of the coin. He has worked for many retailers, then also on the vendor side as well. It would not be a Retailgentic podcast if we didn't cover agentic shopping. Andy shares his point of view and thoughts on where we are today and where agentic commerce is going in the future. Finally, Andy, in his role, Andy talks to hundreds of retail clients a
month and has some really interesting viewpoints he's gathered from the summation of those conversations. Uh, so we get to hear what are retailers thinking about, what's motivating them, and where is agentic commerce on their, their priority list right now. I really enjoyed this conversation with Andy and hope you enjoy it as well Ananda, welcome to the Retailgentic Podcast. Oh my gosh, I'm, I'm excited to be here. Thanks, Scott.
Yeah. You and I, uh, have, uh, we just really officially met. I think we bumped into each other at some trade shows. You've been in the, the industry for quite a while. Um, and I was drawn to you because you wrote a really good paper that we'll talk about. Um, and you're, you're, you know, VP of research at IDC, but before we go into what that means, let's go, let's go back. Let's talk about, uh, I think it's interesting, I learned in a little pre-conversation you have an engineering degree like I do, so we, we share that in common. Not a lot of engineers in the, in the retail world.
So let's, let's start there. What, what, um, what'd you study in school, and then let's, let's draw a path from there to how you ended up pontificating on agentic commerce here today. So I went to Clemson University as a engineering student, electrical engineering. I wanted to get into computers. Was really excited about parallel processing, which is really a cool space in the eng- engineering space. I, uh, graduated there, got a master's in, um, uh, electrical engineering from, uh, University of Massachusetts in Lowell. And then, uh, went off to Northeastern for my MBA later on when I realized
this isn't quite what I wanted to do. The engineering is great for a background, it sets the stage, but I wanted to get into the business side. It was so much more important, and it actually drives the value, uh, for a lot of, uh, especially in the retail side, uh, but in general across the market around applying that engineering into, uh, real-time settings, which with some real dollar values associated with it. So that was the exciting thing about, um, me moving in one direction or another. Yeah. How'd you end up at Clemson of all places? I'm, I'm from South Carolina and I, I'm gonna hazard a guess
you're not from South Carolina. But how did you end up in Clemson, South Carolina? My whole family went there, uh, like five cousins of mine. My aunt and uncle both were professors there of physics and math, so it was all one of those things where family connection that really brought me there. But Clemson's a great football school, it's got great, uh, programs as well in agriculture and other stuff too. So, um, I, I truly enjoyed my time there and, um, you know, of all places, uh, South Carolina was not one of the places that I expected to go. I was thinking about, you know, California, Texas, uh, even,
uh, Northeast, but I wanted to get away from my parents. That was part of it too. So You can, you can do that at Clemson, you get away from everybody. It's just like this little isolated city. Yeah. Yeah. Did you get your MBA straight out of school? Uh, so you did undergrad, masters, MBA, or like, did you go something in the middle? No, no. So, uh, actually it was well Yeah. So I was actually going for a doctorate in, uh, electrical engineering. That was my plan at the time, and I realized that, um, uh, this wasn't exactly what I wanted to do. It was too much on the side of really, uh, getting into really detailed
function and understanding, and it wasn't about, uh, computers anymore. It actually moved into acoustics and other types of, uh, mechanical, electromechanical engineering, which did not, uh, suit me at all, and I actually went and attended a seminar a- about business, and it totally changed my viewpoint on what I wanted to do next. Um, it pushed me towards entrepreneurship, pushed me towards, uh, you know, starting my own firm, building out a, a different type of, uh, uh, focus on where I wanted to go. And it was a, a, I think, a great change in, in the direction.
Um, so a few years later, after I joined a couple of… and did some, uh, some work, I went back to Northeastern as an executive MBA program. So, um, that was, that was true- truly worth it. Very cool. And then, uh, so you finished your MBA, and what was your first job out of, out of that program? Oh, af- after my MBA. Oh, well, yeah, uh, I worked at monster.com. I don't know if you, that means much to you, but everybody's probably heard of it. There's two Monsters. There's recruiting and cables. I'm assuming, uh, you did recruiting.
Yeah, this was, uh, recruiting. This was basically all online and, uh, I was, uh, the coolest part of the job was, uh, I was brought in as a senior manager of, uh, e-commerce and e-business innovation. That was the title actually. So it was really, uh, an exciting space to be. It was at the time when Monster was really growing rapidly. It had some, uh, uh, super people working there. Um, uh, we had, uh, a, a lot of open dollars and availability of resources to build whatever we wanted to build. And, um, uh, you get to see, you know, kind of the intricacies of how,
um, you know, an internal e-commerce site really operates and works. We learned all about, you know, uh, uh, the business analyst side of things, the, the understanding the metrics, all the different pieces that go into job hunting and, and that whole piece of it. I moved at, at Monster a couple ranks up, uh, became a product dir- a product manager, and then a product director there. So that was a load of fun. Um, uh, and I do miss little bits of it, pieces of it, uh, but it, it was also, uh, just a stepping stone to where I wanted to go. Um, I did, and you know, after Monster, I left to, uh, move on to other types of
things, and I started getting into retail. After a few other steppingstones in between, uh, I joined Talbots, got, I got involved with, uh, Staples as well, um, and both all on the e-commerce and, uh, you know, side of things. But it gave me an understanding of what is needed from a digital perspective for a lot of these companies. At the time, they were all starting to go through digital, initial stages of a digital transformation, and that was so critical for retailer at that time, right? I mean, it was such a big change and big stepping, and many retailers hadn't even gone through that yet.
Um, so those were really nice steps to go through and understand the space better. When you were at Talbots and Staples, were you, um- Involved in, you know, that, that's kind of the era of re-platforming. Did, did you get involved in some of those big re-platforming kind of things? Yeah. Yeah, there was a lot of that Yep. There was, there was s- several of those projects that we, we were going through. We were also, you know, at Staples, uh, head of personalization, which was an interesting space to be in because they needed, they needed to kind of figure out what is, what is this whole world of recommendations
and online, uh, engagement all about. Um, uh, it, it gave me a different perspective on what to think about when a company is really looking to connect with the customer. And that was the biggest challenge for a lot of retailers, figuring out what is that digital connection piece that really makes it worth their while. And at the time they really didn't have… I mean, it was just at the beginning and start of the omni-channel revolution, if you will, and where everybody's looking at everything from a single, uh, single perspective. Um, and what I got out of that was, you know, the average customer,
when they're looking for a sweater, they're looking for a sweater. They don't really care how they buy it. It doesn't matter if they're buying it in store or online, through BOPIS, through whatever other tools that they have, BOPIS being buy online, pick up in store, right? So, um, that was a, that was a kind of a, a revelation, if you will, for the industry as a whole. And, and for me, just looking and talking with the customers, engaging at that level, uh, changed my perspective as well, uh, about what retail really needs to think about. It's kind of funny. They're, they're very much… All the omni-channel guys were very resistant to, to marketplaces.
Uh, and then suddenly Walmart and Target ha- helped crack it up, crack it open. But then they've all become obsessed with retail media networks, as you, as you know. And marketplaces are an unlock for that. So, so it's kind of come full circle that the shiny bulb of getting some of these like pure margin ad dollars is bringing them back into the marketplace conversation, a lot of these retailers. Absolutely. Absolutely. I, I think that's, that was a, you know… Retail media networks right now I think are on such a tear. Um, they, you know, uh… I just saw this little article, I think it's Kerry Masters, I believe, who was
writing it, and she's i- in ar- up in arms with another expert in this space, uh, about whether it matters or not, right? Yeah. It's this, uh, annoying guy named Jason Goldberg. Oh, yes, exactly. That's right. I knew, I knew, I knew there was something. He's kind of the grumpy old man of e-commerce. I thought it was a cool argument, though. And, and he, and, you know, Jason makes some good points, but I, I don't agree with him at all. I think, uh, retail media is at its infancy, nascent stages. It's got a long, uh, you know, road ahead, and I think it's gonna permeate
not just retail media, but also AI and other things that are really gonna change how programmatic works, how everything will, will start operating in the future. People are gonna, you know, kind of assume that it's gonna work. They're not gonna even see it happen. It's… That's the cool thing about AI. It's so embedded, right? It's almost invisible, uh, once it's working properly, right? That's where, where I think it's headed. And people aren't even gonna think about AI as something different and new and a shiny object down the road. It's all gonna be part of, oh, well, you know, just like the internet or connecting to something, right?
It's gonna be kind of routine. Uh, back to your bio, and then we'll, we'll definitely wanna talk more about retail media networks. Uh, so you did a stint at retailers, so you were on the, on that side of the house. Uh, and then you did some pontificating. I saw you were at Forrester, and then you did a vendor. You were at Oracle. I, I was at, uh, Forrester for a while, and then I went to, uh, yeah, and I was a retail analyst there. That was fun to do. I enjoyed that work. Uh, Forrester has actually a great team of folks. Uh, not as good as IDC, of course, but they're, they're there. I mean, uh, more importantly, um, uh, the, the thing was, uh, uh, after
that, I got a chance to go into the vendor side, which was really exciting. Uh, Oracle and, and their RGBU, their Retail Global Business Unit. Um, and we had some fantastic people working on that side of the coin. I really got to see how these vendors really opened up the doors, especially a company like Oracle, which has end-to-end solutions, complete, um, you know, opportunities for people to kind of, uh, register different parts of what they need, uh, and then build that into their solutions. Now, there's lots of, uh, uh, opportunities in, in vendors, in the vendor side anyway.
Um, but, uh, Oracle kind of gave me that opening and understanding a little bit better and, and, and the problems that these, uh, companies had with implementation, deployment, rolling out solutions. Um, so that was, that was a great experience as well. That was 2018 to 2020. Um, the… Yeah, so that's post a, a good bit of ATG, right? So they had integrated ATG. They kind, kind of turned it from on-prem into hosted or at least an option to host, that kind of thing. Yeah, everything was all about, at that time, it was all about moving to the cloud, right?
I mean, that was the big play at the time. So that migration to the cloud, it was talked about almost every other call that I was on and, um, you know, how to, how to actually make that work. Uh, I think they did an okay job in pulling that off at the end of the day, but it's a long-term process, right? I mean, every, every one of these retailers that go down this path are putting a lot of money, effort, time, resources into building that out and, you know, some of them just don't quite get there because they didn't plan it out properly. Or they don't know or have the understanding of what's coming, right? I mean, they don't understand what they need to p-prepare for,
the flexibility and agility. So those things, uh, I think are all pieces of that puzzle. Went from there to Diebold Nixdorf, uh, which was a real cool place to work as well. I was there as a, a pricing strategy, uh, guy who, uh, instead of just on the strategy team in general, I was more on the pricing strategy, uh, side of the equation, which was really exciting. Uh, went back to MIT to get a certificate for that, uh, for pricing, which was fun. Um, and then, uh, uh, be- being able to actually apply it in the market was even more exciting because, uh, in the retail sector, that's such
a critical piece of the puzzle. Um, both, uh, you know, merchandising pricing as well as, uh, business-to-business pricing is… Yeah, so it, it was a, a great experience to understand, uh, how Diebold Nixdorf, a point-of-sale vendor, was really driving value across the board. A- and I also got to see a lot of really new, uh, and cool innovations that they were putting out with their, uh, you know, just their general point-of-sale tools. Self-service capabilities that they were working with, and I don't know if you know anybody from there, but folks like Matt Redwood and, and, uh, um, uh,
a few others, uh, that were kind of, uh, driving the show, um, were really driving and betting on self-service to continue to grow, grow across the board. Scale was kind of the, the name of the game at the time. And you get to, uh, go to the bigger booths at NRF when you're, when you're with that vendor. Well, Oracle has one too. Yeah, you're in the… You're playing with the big boys at, at the big show, yeah. But you also have to go on the whole weekend when you're with those guys too. When you're a retailer, you can kind of show up whenever you want, but, like, when you're a vendor at, in the big booth, you gotta be there for the whole weekend.
Um, and then how'd you land at IDC? So, uh, uh, so my former boss, Leslie Hannah, she called me up and said, "Hey, Nada," or, you know, well, I don't know if she called me up or I called her first or whatever, but we talked. I knew Leslie from before 'cause, uh, she was-- she's just a smart person in the space, and she's, she's been with us for such a long time in the, in the industry, and she knows just about everything about it. Um, uh, but, uh, I tried to impress her and did my best to do that. But whatever I did, it seemed to have worked. She decided to bring me on, on her team, which was great.
Uh- There were no other candidates, so you won by default. Maybe, maybe that was it. Won by default. Uh, but in any case, I'm just happy that she, she brought me on the team. And that was the, that was the kind of the s- the starting point. Um, uh, just recently I took over, uh, her role. She retired from, from the, the company just recently. So, um, I kind of, uh, now am, uh, running, running our IDC retail intelligence team, which is, is a blast. It's fun. We have a super group of smart people, uh, in just about every single aspect of the business, uh, when we think about retail.
And what I really like about it is it has a full-fledged team, and it's not just one or two people, uh, taking on retail. They really understand the retail space. Many of them have been there, worked with retailers for so many years. Uh, and, um, uh, you know, that alone gives a, uh, a very different perspective than some of our competitors. And not to dis- you know, uh, uh, to plug, give myself a plug, but I think IDC is one of the best, uh, retail, uh, research groups ever that I've met, that I've engaged. Um, that said, uh, you know, uh, that's just a self plug.
So that's really all that, that really means. But, uh, it is a, it's a good team. Yeah, I do think it's good to get it out there. So like, so retailers mostly, I imagine, you know, in my mind, IDC goes way back. It's been… It's like one of the old, you know, almost like, yeah, pr- precedes Gartner and Giga and For- all these things. Um, and you know, in my mind, it was really tied to the computer industry for a long time, where IDC was kind of like, you know, covering Dell and like the computer industry seems to be-- I think that's my memory of how it was born.
Um, and so I guess your customers are very large, you know, you know, probably tend to be on the larger side. IDC's always been kind of like blue chip, Fortune 500 type kind of thing. So you have the larger retailers, which kind of implies omni-channel, 'cause they tend to be the larger ones. And, and they, they pay you guys a monthly fee to get research and to meet with you, and maybe they're thinking about re-platforming or they want to talk about retail media networks, or do you guys even go into, you know, being, you know, there's probably a physical-- you probably have pr- people on your
team that are good at the point of sale systems and, you know, all these things. Is that, is that how it works? Yeah, absolutely. So w- our team is kind of broken out into different groups. We have Margot and Filippo working on technologies and, uh, strategies inside the store, as well as, uh, you know, experiential operations inside the store. Point of sale systems, for instance, those are all part of, uh, their group. Uh, um, they also cover things like technologies inside the store like, uh, you know, network systems, uh, IT, uh, uh, capabilities, RFID.
All of that is kind of covered on that side Uh, digital pricing kind of falls under my bucket. I'm still also an analyst even though I'm running the team. I'm still active. Yeah. Uh, the tags, uh, so ESL tags, um, okay. So those are kind of me and, uh, Filippo are kind of covering that. I don't want to say it's all me. Uh, yeah, tag- we tag team them, literally. Um, but more importantly, um, uh, I think we have some great people on the supply chain side, too. Uh, Jo- Jordan, uh, Spear. We have Dorothy, who kind of, uh, runs our, um, hospitality, travel, and dining.
And then we have folks on media and entertainment, Alex Holtz, who kind of covers that entire space. Um, so it, it, you know, really a talented team just here in the US. And then we have folks in Europe, uh, uh, Filippo and Ornella both, uh, working out of, uh, uh, Milan and, and UK. Uh, Ornella is covering, um, you know, customer service, customer experience, as well as, uh, uh, CDPs, channels, uh, and also, um, uh, loyalty programs. Uh, and then, um, uh, we have myself, who covers merchandising, marketing,
analytics, and specifically artificial intelligence and retail media. So those are, uh, kind of some of the team. There's some more team members as well in Asia-Pac, but that's, uh, that just rounds out the team. And that's-- the thing about it is it's a team, right? I mean, it's we all kind of interface with, with our customers as needed, as opposed to kind of locking into one particular rep who's-- or one particular analyst who's working with a retailer or a vendor for that matter. Yeah. And retail used to be kind of like a taxonomy. Like, you would talk about topic A or B or C, and then it's quickly become
a graph where it's all interconnected. And, you know, especially being the AI guy, you, you could talk about digital price tags, point-of-sale systems, and, you know, what we're gonna talk about, which is agentic shopping. The, it, yeah, yeah. All in the same conversation. So that's, that's one of the challenges I think a lot of, uh, retailers are just starting to get a hang of. Um, they need to have to make those conversations work. They need to have the data in the same, same way. They need to have their structure, their organizational structure operating the same way. They need to have, um, uh, the resource structure for the
technology to work the same way too. So all of those things make it that much more difficult to kind of fit in to give a holistic view of the retail business. And I think that that's a, a great stepping stone for anybody who's really, uh, who wants to understand and, and get into the AI side of things. You know, when I started AI, it was- wasn't in, you know, when I was doing my first job at Monster or, or anything like that, or even before, uh, when I was in the defense contracting indus- industry. It was when I was in ninth grade, and I wrote a, a, uh, a paper, uh, basically
an essay about artificial intelligence. Those things at the time was all about things like the Turing test and stuff like that. What, what really constitutes AI? Um, what, what does a Bayesian model look like? What's a Markov model chains and stuff? These were things that were, uh, really cool to read about and stuff, but it was almost like a, a future vision, right? A sci-fi vision. You can't actually achieve that in that time period. This is 40 years ago or something like that. I don't know how old I'm now. Don't wanna date myself on the call, but it was a long time ago.
Uh, and the main thing was, uh, you know, the things that I had kind of hoped and wished for at the time when I was writing that paper are coming true now. They're actually happening. You have real live LLMs that can actually do the things that, uh, that were, you know, kind of imagined in those days. Um, and it's almost magical because, uh, you know, it's new, it's technology, it's kind of cutting edge. Uh, and that's what makes it really exciting to be in this space right now. So you were, you were into this since ninth grade. When did…
Did you raise your hand at IDC and say, "I wanna do that?" Or like when ChatGPT dropped, that's kind of-- To me, that's like when this really started. Yeah. Leslie asked me the question, said, "Do you like AI?" I said, "Love it." And I said-- And she said, "Okay, I'm gonna put you in charge of that." That's kinda how it worked, and she was very good about it. But, uh, the thing is, uh, IDC actually has put together a lot of effort behind it, though. We have other groups, uh, that are more horiz- horizontal rather than vertical teams that are dedicated specifically to AI. You know, Ritu Jyoti works a whole team of AI-specific experts that
really understand the space well. Uh, and they understand almost, uh, not just the technology and the capabilities, but from that horizontal perspective across all industries, if you will. Uh, and from technology companies that are really pushing really hard to make sure that their AI gets to the top of the coin. You're talking about companies like OpenAI, Microsoft, Google, Amazon, you name them. Um, they're all trying to push for that. ownership of this space. And I don't think it's gonna be that easy to, uh, to say that, uh, that any one company will really dominate.
Uh, matter of fact, I think it's going to become a mix and then eventually it'll fade out into the points where, um, there are many different players that are kind of, uh, contributing to it. Yep. So you, you probably have, I don't know how many conversations a week, but a lot with big brands and manufacturers. What, what are some of the top of mind topics? Like obviously I think AI, but let, let's dig a little bit, like we mentioned retail media networks. Um, what, what are like maybe the top three or four things that, that are on people's minds these days? If you just go to an NRF con- you know, event, you'll see exactly what's going on.
And they, they basically all talk about AI obviously. Uh, AI is one of the big deal items. There's also retail media networks. I think that is also a very hot topic right now. Um, but a lot of it also goes back to some of the basics, the fundamentals of retail, right? How do you keep a store clean and make sure that people come in the store? How do you, uh, you know, drive traffic to a website? These things are still essential components of retail and a lot of, uh, retailers, they're, they're looking at that. The other piece that I think is really important is, uh, data and data manipulation and I wouldn't call it just manipulation, but also, uh,
leveraging data to, to be able to make decisions and good decisions. Decision-making data, which is a different, uh, you know, context than just collecting data and then having it sit there in some database someplace. It's the actual, uh, action-based data that you can actually use to make good decisions that translate into dollar value for the companies. That is where I think a huge amount of effort is also placed, and that ties back right into retail media, right into AI, right into some of the needs that come from the new technologies that are out there. But yeah, it's still running a business and operating it well.
Let's talk about, um, so one of the reasons that it kind of like spurred me to, to reach out to you was you wrote this amazing paper, um, called "Win-Win Product Discovery: A Retail Agentic AI Use Case." And this is near and dear to my heart because part of my theory here is, uh, so I, I have a theory, so I'm a marketplace guy. I have a theory that, you know, number one, I feel like e-commerce has really slowed down and not really innovated on the user experience. That's theory number one. Theory number two is there's all this activity of people using the, the LLMs.
I think that's gonna spill over into commercial, which is gonna spill over into commerce, and I think they're looking to monetize, and I think commerce is the lowest hanging fruit. When you, you read all the surveys, consumers are increasingly saying, "This thing's helping me research," and then I think they'll increasingly research, find, and buy products. Um, and then I think the other thing I think a lot about is, and I learned this at Channelyzer, is our, our product discovery engines are not great because, um, we, we have the lowest common denominator problem that, uh,
goes all the way down the, the whole tech stack to the product catalog. It's, it's kind of, you know, the way it's put together from different systems, the attributes it has and lacks are gonna be a bit of a, uh, a challenge for retailers. So, so you're, you're, you're… Talk a little bit about the, what led you to write this white paper and some of the things you discovered or recommendations you're making. And yet I also discovered, I also learned from this paper you, uh, you either play a lot of baseball or you have sons that play baseball 'cause, uh, you have a, you go really-- I was also shocked how deep you go.
You go, like, way into databases and retrieval. You go deep. Uh, so, so that's when I was kinda like, "This guy has to be, like, an engineer or something because it's, this is not surface level stuff." Yeah. Yeah. Well, the, uh, so the engineering, um, background definitely helped a-and, and I appreciate that, and thank you for the compliments about the report. I think it, um, more importantly, uh, it came from above saying, you know, "We need something that's deeper than just that s- you know, top level stuff. We need something that really goes into how these things are working, what is the real value add that they're bringing to the table, and then the benefits of
bringing in understanding of that space." I think that was w- uh, the, the question that was asked from one of my senior leaders at our company. Um- And, um, uh, more importantly, I think the, the second piece to that was how do you apply it to, uh, a vertical, right? That was the question that was being asked at, at this company meeting. And, um, uh, so I said, "You know, there's some key use cases we've already put together for gen AI. Why don't we see what-- how that fits in for agentic AI too?" And that's really what kind of spurred me on to write the paper.
Um, uh, it was also because I think agentic AI is a, uh, is something that is gonna be overtaking the market in the, in the future. I mean, without a doubt, it, it will dominate eventually. And the problem for most people to under- is the understanding of what it means, right? What is agentic AI in the first place? What is that concept of an agent, um, uh, really doing the work for you? Uh, and I think there's some key pieces in there that I talk about, such as, um, you know, reasoning and, and thinking engines. What do those really mean?
What does that really, uh, uh, you know, compose of? Um, the ability to con-constrain those engines with RAG, for instance, or retrieval augmented generation, for instance, or, um, uh, build-building in, uh, capabilities that actually take it to the next level of where you have multi agents or a sequence of agents that are actually doing multiple problems at a time to come up with, "Here's the sequence of what you need to do." And lastly, probably, uh, orchestration. That's a key component to the, to the whole concept of agentic AI, where you're orchestrating several different pieces together, and the
tool is doing that orchestrating. It's just like it's doing, uh, you know, the probability-based, uh, uh, estimation of, uh, the next sentence it's gonna talk back to you about whatever topic you ask it. Uh, it has to do the same thing for here's the sequence of actions that you need to take to accomplish this goal or whatever that task might be. Um, so, uh, and, and I should probably add one more thing, and then it's the execution, right? It has to be connected to something that's then running that program and executing on. So once you have all those pieces- So, uh, and that makes it a real function as opposed to just a, a theoretical, uh, a theoretical one.
And I think, um, uh, uh, that was what I tried to, uh, explain here and to show in the form of a use case, and I took this one use case here. Uh, but I'm gonna-- we're gonna see this happen across many different retail use cases down the road, not just one or two. Even our gen AI taxonomy that we've developed, um, really covers something like, uh, you know, 20 different use cases that are, are primary, uh, uh, use cases out there. And to your point, uh, Scott, a lot of them are in the space of digital eCommerce product, uh, you know, uh, descriptions and, and similar, uh, that
are, uh, in many cases, productivity related right now and efficiency related, but in the future, they will be revenue and sales growth related, right? It's gonna change and shift in time, uh, towards something far more, um, important to the retailer I shouldn't say important. I think cost saving is just as important, but, uh, you're gonna see it become more, uh, real for, from a dollar and bottom line point of view, um, from profitability point of view. So that's, that's gonna be cool to see. So when you're talking about agents, sometimes I, I hear people kind of use these interchangeably, and I think as an industry we need a little clarification.
I kinda think there's, there's kind of meta agents that are gonna look across sites, and then there's like intra agents that are gonna look inside a site. So Rufus and Sparky are examples of, of agentic shopping that lives, you know, they're within the, the walled garden of an Amazon or a Walmart. Um, and then you all… You know, there's another one that's kind of more like, you know, Perplexity Buy with Pro, which is looking across all retailers. Do you, do you… Or are you, are you generally talking about either of those, or are you specifically talking about retailers should be
implementing the Sparky-like thing? Or what, what's your point of view on that? There's, there's conditions and different use cases for both. Yeah. And both of those things will be adopted over time by retailers. You can't just stick with one or the other. Uh, sometimes you need to have that constraining structure behind, uh, what you're doing. You need to have the, the blocks to, to get answers quickly and to have more relevant answers anyway to whatever you're asking for, uh, whether it's execution, whether it's actual question, uh, a generic question that you're asking. Uh, so Rufus, Sparky, these guys are gonna be very important, but just as much as,
uh, you know, all the models that come with Perplexity, and they've, they've worked with Claude, with Gemini, with… You know, they, they got all of these different models already built in. Um, ChatGPT of course. And, and so you'll see, see a lot of that as well. I think, um, uh, the example I wanted to give here was where do you start? Kind of here's one example case that makes sense. Um, it's, it's a, a fundamental case for retail anyway, which is product discovery. How do you actually go about finding the products you want? Uh, and then what are the tools and capabilities that help you as a consumer in some cases, and then also the retailer?
One of the interesting things about this, um, this particular use case is it's a win-win for both the retailer and the consumer. Consumer can find the things that they want faster and more effectively, uh, and the retailer is able to push and upsell and cross-sell items that are more relevant to the consumer as well, um, allowing for both, uh, parties to become, uh, you know, uh, get that, uh, value add at the end of the day. I think that's an important component to keep in mind, uh, when they're thinking about what types of stuff they wanna build out. If it's just, uh, you know- value for one group or the other, it's, um, it's gonna
have some challenges in becoming adopted, becoming, uh, you know, uh, driven as a, as a major value add across the board. Yeah. Let's, let's dig into that. That, that's interesting 'cause a lot of the retailers I've talked to, there's a fear that they're gonna be disintermediate- disintermediated, right? So this… The-- they want a, they want a human to come to the website and then upsell them stuff, and you're, you're kind of making it sound like if you can get good at your agent talking to the, the buyer's agent, then there still will be upsells. Is that, uh, kinda what you think is gonna happen? Actually, I think in the future, the agents are not gonna be, uh,
recognizable, uh, from a human. You know, they're gonna have almost the same capabilities. They're gonna have almost the same figuring-- ability to figure out something like, "Oh, that's a promotion. Maybe I should take advantage of that now because it's, you know, it's up there for 15 seconds," or whatever amount of time they're gonna give you. Um, because, uh, my, my human overlord, uh, is saying that they want the best, you know, product at the lowest possible cost, right? So, uh, there, there's gonna be a lot of that as well, uh, and that's gonna take experimentation. That's gonna take, you know, exploration of, of these places, spaces.
But there's no doubt that those things are gonna evolve over time, uh, to become more human-like. I think that's just inherent to, uh, the nature of our business. Do you see a day where agentic traffic exceeds human traffic? And it's like, can you… Do you have, do you have a prediction? Is that like, are we 12 months away from that or like five years? Oh, I don't have a date, but I, I know it's gonna happen. It's, uh… And it'll happen sooner than later. We're-- We don't know right now, but because it's an exponential trend, uh, I think it'll be adopted very quickly. People will start seeing more and more of this- All they need is one,
just like ChatGPT shot through the roof, uh, in November 20th, 2022. Um, same thing is gonna happen here. You're just gonna see one or two examples or one, two use cases, and then people will start adopting like never before. Oh, there's my agent. I've got one. I got my own personal agent. People with more money will, uh, you know, have their own fleets of agents and, and they'll use them. I mean, it'll be a lot easier for them to go out and, "Oh, I love that coat that that celebrity was wearing online," you know. "Let me, let me see where I can get that for a low cost," right?
Click, there you go. A few days later, like… Or maybe not days, maybe the same day the, the coat is arriving at your door, right? So that's the kind of thing that I think is, uh, is, is gonna be inherent to the mo- models of, of the future. Um, but that's still a ways away. We still have some work to do to get that working to that level. But, uh, yeah, it's all gonna be, it's all about human nature. We're all greedy for stuff and we're still materialistic in a lot of ways. When retailers-- I'm sure this report stimulated a lot of discussions. When retailers say, "Okay, I get it.
This, these shopping agents are gonna be coming to my site. What should I do today?" What, what's your… Do you have some tactical recommendations? What they have to do, start thinking about is, um, uh, setting up sites that are able to cater to both humans as well as non-humans, right? I mean, as well as to agents for that matter. Uh, and, uh, that doesn't mean, uh, necessarily changing everything nes- uh, you know, from the start, but it has to be easy for, uh, agents and humans to be able to operate on those. Um, it may still use the same types of stuff. The whole, uh, concept of gen AI and, and, you know, that I think
is really, uh, has been the major change is that it now understands things at the human level, right? So, um, an agent should be able to do that too. The only distinction there is it'll be able to do it faster, right? It'll be able to figure out stuff in two seconds other, rather than 30 seconds that it takes for a human being to con- you know, conceptualize a page and figure out what's on it and, and do what- whatever's, uh, you know, uh, associated with it. So I think, um, uh, you're gonna be able-- You're gonna have to have tools that are able to capture it very quickly.
You'll have to have real-time capabilities. You'll have to have, um, the ability to think like a human first and then like a hu- and an agent second. Agents are still have, maybe may have one or two purpose points. Uh, and the initial set of agents are still gonna be a little slower at being able to capture, let's say, uh, sentiment or, um, uh, what a picture might be representing for, for a, a juicy hamburger on the screen or something like that, right? But, um, ultimately those agents are gonna understand that too. I mean, that's not gonna be difficult for or challenging for them in the not-too-distant future.
Okay. So you need, you know… What, what, w- we're having conversations with retailers too, and you know, what, what we have found is looking at the current agents that are out there, they, they love context. It's just like that's their- You know, context window and everything. So the more attributes you can give them, like you normally wouldn't show… Y-you know this being an engineering person, right? Like, you're not gonna show someone the, you, you're gonna give them… Humans are looking maybe for three or four bullet points, but the LLM wants like 1,000 data points. You know, it, the more it can learn about the product and it, you know,
it's makes, it's life, life easier if they could just all be on the PDP. So they're, you know, the… W- I think one way it's gonna be different is the human PDP is gonna be this beautiful, very imagey thing, and I think the other one's gonna look like a spec sheet basically. That, you know, I think the LLM would like much prefer a spec sheet. Like, you know, give me some images and stuff, but any content and image, just like give it to me in text and, and I'll, I'll gobble that up, and I'll be able to make recommendations up. What do you think about that? Yeah, that's a, that's a good theory. I think that's how it'll start for sure, but I don't think that's how it'll end up.
I think, uh, what'll happen is, um, uh, and, and here's the reason why. The agent will go and find that information elsewhere, right? They're not gonna look at your site necessarily to find that, the specs on that particular product or how well people adopt it. They'll find stuff and data points that are associated with that product that you never even knew were associated with that product. Um, an example might simply be I'm going to the site, I see, uh, let's take baseball bats for s- for instance. I, I have 1,000 baseball bats. Here's, uh, uh… It's gonna go out and find out which baseball players, Major
League baseball players actually use that bat, and it'll come back and use that as a crit- criteria now- Yeah to identify which bat is best for this particular client, and it doesn't need to do that from your site. It'll actually do it on its own. It'll figure out how to do it on, on its own and where to find that information on its own. Good point. So I think that's really where it'll eventually get to. Yeah. Uh, but to start off, yes, absolutely. The spec sheet is important. It's gonna be, uh, it'll make it easier for, um, uh, the retailer… I'm sorry, uh, whoever's using it. It's not just retailers. I think it'll be beyond that. Uh, and more importantly, it, it's around the metadata, right?
It's all about that metadata set that is describing what you're looking at, that context that you just mentioned here. I think that's such a cr- key component, um- Because it's the main piece that's, uh, been eluding much of our past technology. Um, and, uh, we just haven't been able to capture that. Now that we can and we have the compute power to, compute power to do it, uh, it changes the game completely. Since you're the broader AI guy, I could talk about this agentic shopping forever, but let, let's mix it up a little bit. Um, I think it's interesting too to think through, you know, so, so if
you think about the foundation retail is built on, there's this product catalog and inventory and supply chain. There's also, you've talked about pricing, you obviously have a passion for that. Um, and then there's like loyalty and, and you mentioned that, uh, as we were chatting, um, and personalization. In, in a-- You know, it's gonna be interesting, do you think those, those elements of the foundation have an agentic flavor and a human flavor as well? Because you could see this agent-agent protocols being developed that, you know, to your point, you say maybe your agent says, "Hey, I know," you
know, maybe part of your prompt is, "I kind of need this baseball bat today." So that's gives me the context that Andy's son has a game today and he broke a bat, now he needs a new one or whatever. You know, and, and maybe I make you a special offer if you buy in the next five minute. So there's gonna be… But then on, on the retailer side, there's gonna be this calculus of, you know, maybe I'm swimming in this one baseball bat and I say, "If you buy this one Easton bat, I'll, I, I'd be happy to give you 10% off." It's a little bit, you know, you, you told me $100 price point, I'm gonna get you to $150 with a discount and maybe get a little bit more out of
you because I'm swimming in this. I've got a lot of inventory on this bat or a better margin or whatever. Do you-- Yeah. Is that, is that a… How are all, how are all, some of these things gonna play out, you think? Wow. That's a, that's a deep question there, Scott. You asked me to predict the future for the next, uh- That's your job. You're, uh, you're… It says there ri- retail insights. It's right there in your title. That doesn't mean I know it all. Oh, we know you don't know it all. That's the whole fun part of it, like- Just this is my job. So what I, I, I'll give you my thinking on it, and, and, and you
can take it the way you want, right? I mean, maybe that's, uh… So I'll, I'll give you my thinking on it. So I think there's going to be a, um, uh, so yes, there is that connectivity between the agent and agent, uh, onga- ongoing, and I think that's gonna become more and more, um, uh, important over time. There's gonna be, um, uh, agents of agents, if you will. Literally an agent who's operating on the behalf of several other agents in the, in the, in the background who's doing their task, and it's not gonna be kind of like an orchestration agent, but an agent who's just like a travel agent
would on beha- on somebody else's behalf. Um, the same type of functionality, but it's on behalf of other agents. So the overlord's no longer human, they're an agent of some sort. So I think that's gonna also be, uh, the kind of that multi-structure, if you will, of, uh, being able to engage. What I think is gonna happen over time though is, uh, there's gonna be a level of authenticity, right? You're gonna have to find out where is that human in that loop, and, uh, how, uh, you know, and put more prioritization and, and engagement with that human side of it wherever they are in that process.
Whether they're the ones who say, "Oh, go find me a bat," or whether they're the ones say, um, uh, that ha- kind of gave the initial instruction of, uh, "Let's get this done in a day," or whether they have a problem that they just don't know how to solve, and the agent is then re- representing that, uh, that particular human in trying to get something in a day, the broken bat or whatever example you give. So I think, um, we're gonna see a lot of, uh, of, uh, experimentation up to the point where, um, then there, uh, there will be eventually a prioritization of what's most important versus what's second most important, third, et cetera,
and they have to figure that out. The, the agents themselves will do that. We had a problem like this before with IoT, if you remember. Uh, we've been talking about that in the past, right? Where, uh, you know, all these internet c- you know, computers would be talking to each other, and there is some of that that goes on, uh, for sure. But it didn't expand and explode like everybody thought it would. I think agents are gonna be a little bit different. There will be a little bit of an explosion there, and people will devise, uh, tools to do some of that. And to the point of, uh, you made before where, uh, you know, does this fit into pricing? Does it fit into loyalty?
Does it fit into personalization and stuff? I think that is very much where you'll see different use- types of use cases come out, um, that, uh, literally will look for, and pricing is a great example of that. You'll see many different variations of that in the marketplace, and they're gonna test out these models on their own. So you're gonna have agent- You know, groups of agents that are working together, uh, to kind of find ways of, uh, of getting the best opportunity for the consumer as well as for the, the retailers as well. Yeah.
We could see group buying actually become a thing, but it's gonna be agentic. Yeah. Interesting. Yeah. That's pretty cool to see. I mean, it's, it's, it's gonna change, uh, how we think about doing business, how we think about shopping too. I mean, that's a, it's a big change. Have you, uh, I didn't see it in the paper, and this may be off-script, but did you, have you looked at MCP and A2A and some of those, those, those protocols? I, I have looked at MCP. I think that's a… Yeah, I've seen some of the standards around that. I think that's a, that's a great step, I think, in the right direction. Um, you know, uh, just because it's, it's starting to establish
a standard, but I think those standards are gonna change too. I mean, just because of how, uh, you know, agents are gonna, uh, are gonna be developed. Matter of fact, I think in the not too distant future, uh, you're gonna have agents develop the new standards for that matter, uh, as opposed to what we do today, where we have a broad scale scope. And, and those agents are gonna kinda condense it to just the things that they need at the time, right? It's, it's gonna be very much unique to that particular agent's particular, uh, needs that they have at that moment, and that's how they're gonna actually function. It's, it's a very, uh, interesting world in that sense.
Um, I mean, just, uh, you know, two days ago, just I think two or three days ago, um, uh, uh, Google came out with their Google IO conference, and they had a whole bunch of stuff going on. Microsoft had another conference going on at the same time. All of this stuff is now tying back to retail and what we're doing in the shopping world, and that's a very big deal. Um, uh, Microsoft's Copilot is really an interesting play, uh, because it's now no longer, uh, that coded environment that you have for, for developers anymore. It's just, "Oh, tell me what I want," and, and there you go.
You got this program and code that can do, do exactly what you want. I wanna go scrape 1,000, uh, websites for price points to compare it to what my price point is. Here you go, uh, and, and just by asking the question. And that's the kind of thing that I think is gonna start changing the work environments for, uh, retail, uh, employees as well as other businesses as well. But m- mostly you're gonna see a lot of the impact at the retail side, just because that has the widest reaching impact overall. Yeah. I was excited to see Copilot put out product cards. So we're, we're starting to see the, the intersection of what you and I would call
comparison shopping with the agent, with, with the LLM, which is the first big step. So now that it understands product catalog and buyer intent, then we're gonna get into that agentic flow. Whereas Google went all the way in. They, they went from product cards to buy for me. Um, so they're already… They, they're, they're kind of almost at the furthest, you know, out, you know, uh, of the maturation scale of this. Uh, I don't wanna say that. I, I, I've seen, you know, a lot of these, like I said, Microsoft, Amazon, these guys are also doing just as much.
If you look at some of the stuff that they're working on now, they have almost the same level of capabilities that they're building out. Um, uh, you know, uh, and I think that those things are gonna be the real, uh, real drivers in, in the long term. I don't think there's gonna be one owner, let's put it that way. I think there's gonna be many different players in this space, and the agents are gonna be unique and almost as many, uh, different types of agents as you have different types of models, right? And you can expand on that as well if you want. Uh, you know, not just models, but also how those models are implemented and
applied to different types of use cases. So, um, there, there's a wide world ahead of us. I mean, that, that's the cool thing about it. Um, and a lot of opportunities too. Sounds like job security for analysts. So that's good. Lots of, lots of change and chaos then, uh, it'll be good for you guys. Some people are saying things like we're gonna be replaced by these AI agents too so I don't know. I don't know how much job security there is, but definitely there's a, there's a job security for those who are working in the analyst space and working with agentic AI. They definitely have some job security coming up, uh, at least for a, a decade.
So… Well, awesome. I, uh, I promised we wouldn't go, uh, too much past your time there, so I wanna be, uh, I wanna honor your time. I know you're a busy, busy person. So this is good. We-- I definitely wanna have you back on in like, you know, three to six months, so to kinda keep getting a pulse or if you have any other awesome papers that you write, uh, I think that'd be great. If, if folks wanna follow you, where do you pontificate? Um, so you can certainly go to idc.com and, and click on analysts and type in my name and stuff. That's the easiest way, uh, probably. But, um, uh, just look me up online, you'll find my name somewhere out there.
I got a bunch of different reports coming out. Yeah. Um, and, um, uh, you know, if you go to the IDC retail section and, and look at some of the other players in the market, not just myself, but all of us talk about and think about AI. Even though I'm kind of focused on that, um, every single one of our analysts really focus a lot on AI today, uh, and in- including agentic AI. So, um, uh, great place to connect with us, and appreciate this, uh, opportunity to talk with you, Scott. I mean, this was great. Absolutely. And I, I'll give a personal shout-out. Um, your LinkedIn feed is very good.
So I, you know, if you're gonna … I'm sure, I'm sure the IDC path is awesome, but I've, I've enjoyed your, your LinkedIn. 'Cause you, you give us some snippets of what you're, what you're putting out there and, and that's been a, been a great read as well. Yeah, absolutely. I appreciate that. Yeah. Thanks for coming on. Appreciate it. And I'm sure, uh, we'll have you back on, or I'll see you at one of the trade shows here soon. Oh, I'm sure we will. I can't wait. Have a good one, Scott. Follow, rate, and share to stay up to date with RetailGenic, your guide to the future of AI shopping agents.
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...
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...
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...
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...
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...