From Warehouses to Robot Shoppers: Jason Goldberg Talks Retail’s AI Makeover - Ep. 286

Foreign.

0:00

Welcome to the Nvidia AI Podcast. I'm Noah Kravitz. We're talking retail, the state of AI in retail, specifically with Jason Goldberg. Jason, better known as retail geek, is Chief Commerce Strategy Officer at Publicis Group Publishing. Apologies to the French, I probably just mangled that.

0:10

No, you nailed it.

0:27

Perfect. And he's a well known and well followed expert in all things retail, e commerce, and digital transformation. Jason's also a podcaster, so I'm going to give a quick shout here because, you know, we got to take care of one another. You can check out the Jason and Scott show as well. But first, let's talk AI in retail. Jason, welcome to the Nvidia AI podcast. Thanks so much for taking the time.

0:29

Oh my gosh, Noah, thanks so much for having me.

0:50

So let's dive right into it. AI changing lots of things, including the way we shop. How is AI affecting the way people shop both in stores and online right now?

0:52

Yeah, great question. And first of all, I should acknowledge there's not universal agreement in our little pool of retail. There's a lot of people that think this has been the most transformative, biggest disruption, hugest change in our lifetime. And I, I tend to lean slightly in that direction. But there are a lot of equally smart people that talk a lot about how it's likely overhyped and hasn't had a really big impact and that it might all be one, one giant hallucination. I just want to acknowledge there's a fun dispute about that in the space at the moment.

1:02

I feel like I've heard that same dispute in other spaces as well, so. Well taken and you know, we'll proceed with that in mind.

1:37

Fair enough. So that being said, specifically in retail, I like to think about there being two big branches of impact from AI. The first branch is what I'll call the optimization branch. There's a lot of things we've been doing in retail forever. Commerce is 6,000 years old. We have a lot of processes that haven't changed an awful lot over time. And AI has made many of those processes more efficient. It's made us better at those processes. So we either get better outcomes from the same effort or we get the, the same outcome from, from less effort. And so these are things like supply chain optimization, conversion optimization, labor optimization, and we, we could go get into a bunch of specific examples of retailers that have gotten better at the things they always do as a result of the, all these new capabilities that AI has unleashed on us. And, and that alone would be huge and disruptive. And, and I would argue those are the kinds of things that we've seen already start to happen at reasonable scale. So if you say, like, what is the financial impact been to date of AI? It's, it's largely things in that branch.

1:45

Right, right.

2:56

There is to me a potentially more interesting, bigger disruption, which is how does it change the brain of the consumer and the actual path to purchase? And, and how do we go shopping differently in the future than we did over the last 200 years as a result of AI? And I would argue that while that's super interesting and we can point to examples of where that's sort of hypothetically happening and how it might happen, it probably hasn't happened at scale yet. So, you know, this is January, January in retail is the big National Retail Federation big show in, in New York. And, and you know, this conversation and talking about, about this was, you know, probably 90% of the, the, the energy there. If you were really trying to say what has the economic impact been? Probably not huge yet, but it has the potential to be, you know, much huger. And, and the way to think about this is, you know, what are the purchases that you have always had to do for your family that have required your bandwidth and effort that you might just simply not do that you might just outsource to a robot to make for you? And you know, I, I talk a lot with my retail clients about Personas like the Garcia is the busy family. And the irony is, of course, the Personas that many of my retail clients are worried about tomorrow is the Nvidia microchip.

2:57

Right. So can we kind of ground things for the listener maybe? And you started do that a little bit. And you know, I'm asking this now before I dive into asking you, like, well, wait a second, is this AI shift really bigger than the Internet was for E Commerce? And we'll come back to that because in your, in your intro, that's the first thing I thought of. But can you give maybe a concrete, everyday example of how AI is already changing things for the consumers, perhaps in the way of personalizing shopping or if there's another, better example?

4:22

Yeah, yeah, so, so one, one super simple one is that the two largest E commerce retailers in the United States of America are AM and Walmart. And you know, since we launched E Commerce in about 1994, the paradigm has been pretty similar. We've had a search box, you type stuff into the search box, you get search results, you click on one of those results, you get to a product Detail page. Hopefully you add it to cart and checkout. That search box has largely been the path that most consumers use. There are also navigation menus, but it turns out for most retailers, the only a small minority of consumers use those. The on site searches is super important.

4:54

Yeah.

5:35

And this holiday season when consumers went to Walmart and Amazon, in addition to that traditional on site search, they had an entirely new amenity. They had Rufus at Amazon and they had Sparky at Walmart. These are AI agents that you could type questions into that, help answer those questions and recommend products that met those questions. So in the old world where you kind of had to do the research, you had to do, have all the subject matter expertise and you could get very sort of brain dead answers from the on site search to help you do your own research, you can now outsource that research to the robot. So you know, say you're, it's I live in Chicago, it gets quite cold in the winter here. So the people smarter than me all plan vacations to warm weather places where you might want to wear sunscreen. Right. In the old days you'd have to know what criteria were important to you about sunscreen. Like I would have to already know I need a water resistant sunscreen and I need one that blocks UVB rays and I have a 10 year old. So here's the specific requirements for children's sunscreen and oh by the way, where you're planning on going is Hawaii and you have to use a reef safe sunscreen there. And so I would have to know all those criteria. And then I could type into Amazon those criteria and hope it would give me a selection that kind of met those criteria. And dirty secret, it's actually never worked that well. But today I could go to Sparky and I could say, hey, I'm going on vacation with my 10 year old son in Hawaii and I need a sunscreen. And Walmart would say, oh, here's the criteria you should think about. Here's the sunscreens that best meet that criteria. Here's three you could order right now that will get here in time for your vacation. Right? And so all of that get up to speed and research that I used to have to do for everything, for baby strollers, for car seats, for, for dinner meal planning, for, for ski equipment, for podcast equipment, whatever, whatever search you were doing, you had to become a subject matter expert. And then you could use that on site search to find the right subset of the billion products that Amazon now offers on their website. And the robot, you know, Potentially makes that a lot easier. And while a minority of people use those, those robots, those, those agents, Sparky and Rufus, this year, those that did spent a lot more and had a lot better outcomes. Right. So the conversion rates from those, those on site agents is something like in the early indications we have so far, something like 3x the conversion rate when a consumer use that on site search.

5:35

Yeah, you read my mind. Because as you were describing, sort of the old way of doing things, to put it that way, I was thinking, yeah, that's me. I over research everything and I only go on to the site when I'm ready to type in, I want this product from this brand at this size, et cetera, et cetera. So I haven't used, I'm familiar with Rufus a little more than Sparky. Haven't used them that much, but it sounds like they work. People like them or at least are, you know, they're converting.

8:13

Yeah. And again, I, in my mind, this is the first holiday where they even were exposed to it. Whether they used it or not, they saw it on their path to purchase. If some used, the early adopters used it. And then, you know, at their New Year's Eve parties, they were talking to their slower adopter friends and said, oh man, I did my Christmas shopping with Rufus. And so now those people are maybe a little more likely to try it. Right. And so that word of mouth spreads. And so whereas this may have been the holiday season where we kind of first offered the amenity to consumers, if I were a betting person, I would bet next holiday season is the one where this really starts to happen in very high volume.

8:41

Right. And this is, you know, maybe a little bit of a faux naive question, if you will, that I might know the answer to.

9:19

When I think of naive, I think of you, Noah.

9:26

Thank you. Yeah, my children do as well. Why this season, you know, generative AI kind of hit the mainstream a couple of years ago. And I mean, again, the things you were talking about in the retail space echo to me and, you know, other spaces as well. Is it just that the technology finally felt mature enough to, you know, launch to the public or what was the motivation behind, you know, pushing things out now?

9:28

Yeah, well, so everything is evolving very quickly. Like there's this construct that the smart folks at Gardner invented a long time ago called the Hype cycle. Yes, and I'm a disciple of the Hype Cycle. I feel like in general, every time something new comes out, it starts out by being overhyped. We have even higher expectations for the Benefit it will give than it will actually give. And those heightened expectations result in disappointment. The initial versions of the technology rarely can live up to that hype. So you get this thing that Gardner comically named the trough of disillusionment, as folks don't get as much benefit as they expected from some new technology. But, but eventually the technology gets good enough that it really does deliver on what were the reasonable upfront expectations. So you sort of reach this plateau of productivity, as Gardner calls it, right? And so you think about all the innovations that have happened in our lifetime in retail. And side note, I mentioned earlier, commerce is 6,000 years old. I would actually argue there aren't that many disruptions in the history of commerce. That why I took a really strict definition of what disruption meant and I did a quick count and there's like six or seven disruptions over that 6,000 years where shopping fundamentally changed. Arguably you and I have lived through three, which is crazy, right? Like, so digital, like we used, we used to exclusively go to brick and mortar stores. Today, 25% of all purchases in the United States are online, right? So, so digital was a huge disruption. I launched my first e commerce site in 1994. No one was buying anything online in 1994, right? So it was me talking to like a few friends about some hypothetical overhyped benefit that never happened. And everyone always laughed at us, hey, Jason, no one's ever going to buy a TV online. No one's ever going to buy a movie online, no one's ever going to buy clothes online. And all of this, no one's ever going to buy things have come to pass, right?

9:53

Yes.

11:46

And so then that spawned a whole digital industry, all of whom got wildly disrupted by this new device called the mobile phone. Right? And we all built our experiences around websites and huge monitors. And while we were, you know, and we spent millions of dollars on our, on our capability to serve a webpage. And then ironically, we, we all like spent $10,000 a year on a tool to launch a mobile website, right? And you know, today more than half of all e commerce traffic is on mobile devices. So, you know, I, you would argue mobile disrupted digital. We used to discover all these products at retailers, websites, right? Like that's where you went to go shopping. It was the digital equivalent of the mall is, is the marketplace like Amazon or ebay or Alibaba. And that's where you went to discover things. That is not where you go to discover things today. Because we had another disruption, social, right? And the fastest growing retailer in the history of mankind is TikTok shop. Right? Like we're discovering these things while we're doom strolling at 2am on TikTok, not when we go to Target and browse for new things. So what was common about all three of those disruption is they happen slowly enough that we all had time to adapt to them. Right. So we started talking about them and we started building solutions probably before the consumer was ready for those solutions.

11:46

Right.

13:06

The thing that's been utterly unique and hugely challenging about AI is the pace of innovation. Right. The gap between when this capability was first hypothetically proposed and when it was highly functional in the wild has been way narrower. And I'm haunted by this phrase that some much smarter person than me uttered. The pace of change has never been this, this fast before and it will never be this slow again. Is, is, you know, we're all, we're all struggling to adapt to this. So you, you asked the great question, like, why is this the year? It's, it's because stuff got better so much faster. Like the. Was. Was there LLM agents a year ago? There were, but they weren't reliable enough that retailers had deployed them on their site. So they, you know, earlier this year the, these big retailers invested in the technologies and went live and put, you know, even marketing behind them. You know, maybe they were, they were pilots that were kind of hidden before, but now they're front and center on the homepage of Walmart and Amazon because presumably those retailers do have good confidence in them.

13:06

Right? Yeah. Well, I think ChatGPT, the original launch was supposed to just be a little, a little test run or something like that. Right. And we've seen what's happened. The latest state of AI and retail and CPG survey, which just launched in early January. And listeners, you can download it, read it for yourselves on the Nvidia website. We'll drop that URL at the end of the episode. Found that 91% of respondents are actively using or assessing AI and 90% plan to increase AI budgets in 2026 this year. Tracks along with what you've been saying. Most companies also say that AI is helping them grow revenue and cut costs, which tracks. But maybe you could give some examples of how that could work in retail.

14:14

Yeah, yeah, yeah. Well, so going back to my earlier conversation about these two branches, I think those real like grow revenue and cut costs in meaningful ways have been the efficiency plays. Right?

14:56

Yeah.

15:06

And so last year Walmart, I mentioned Sparky that they launched, that's actually one of four super agents they launched. So Sparky is targeted at shoppers and it's helped shoppers find stuff. They also launched Marty, which is focused at helping vendors sell stuff on Walmart's marketplace. Right, right. And so for marketplaces to be successful, it turns out one of the things you most need is a lot of sellers and a lot of products for sale. And if you look over the last several years, Walmart has ramped up from the 120,000 items they sell in a superstore to 400 million items they sell on their marketplace today. And part of the reason that happened is because AI capabilities of agents like Marty have made it way easier to onboard sellers and list more products for sale. They made it wildly more efficient. In the old days, the way I would add more sellers is I would have to hire more salespeople to do all the manual data entry. But today, the robots are way better at that. Right. They have another agent which doesn't have a cutesy name for the Walmart associates. Right? And so that's made those associates wildly more efficient. So, you know, instead of them having to run around and answer a complicated question that a customer has, they now get a prompt answer from their agent, which allows them to be more efficient. It takes a lot of the drudgery out of working in retail stores. And one of the big drudgeries in retail is scheduling. If you're Walmart and you've got, you know, millions of associates worldwide, like 1 out of 300 people in the US works for Walmart, Scheduling shifts is a huge.

15:06

It's gotta be a nightmare, huge effort.

16:44

And then every one of those associates wants to trade shifts. Sure, they want to, want to change something. And so all of the overhead in Detroit is for those. And they all have questions about their benefits and where their paycheck is and all of these things, all of that has been wildly optimized by AI, making these companies much more efficient. And then the lowest hanging fruit of all these big retailers are also secretly technology companies. Right? And so certainly nobody's surprised to hear that Amazon is a technology company with tens of thousands of engineers. Walmart has 10,000 engineers. And AI has not accelerated anything more than it has accelerated the efficiency of developing software. Right? And so imagine all these legacy systems that any retailers on. I live through this thing that your listeners are too young to know about, called Y2K, where we went from 1999 to 2000, but all the programmers wrote bad software in COBOL that only had two digits for the date. And so we thought the world was going to end and we had to wheel in all of these retired COBOL programmers back to the office to update all their code to have four digit dates. That probably happens every month. Now JavaScript gets depreciated and there's a new JavaScript, there's a new library that gets rolled out and the old one gets depreciated and no one even looks for a programmer that knew how to write it anymore. Today you tell the robot to update all your code to the new JavaScript library and it magically happens more efficiently and better. And I've seen Amazon quote something like hundreds of thousands of man hours that they had taken out of their program. I've heard a quote from Walmart that they've taken like 100 million tasks out of Sam's Club stores associates jobs as a result of all of these efficiencies. So I think those are the kind of the main efficiencies we've seen today. And then, you know, there are like increasingly efficiencies helping people discover products and buy stuff more. So one of the most successful retailers in the world today is she in an apparel retailer. They use AI to scrape social media and decide what clothes to make. And they make the clothes in a few weeks instead of six months, six to 18 months. So that AI that helps them decide what to make means they sell everything they make where a traditional apparel retailer is lucky to sell half of what they make, at least at full price. Right. And now we have somewhat traditional retailers like Zara saying, hey, we're using the robots to optimize our supply chain and our forecasting, and we used to only be able to sell 60% of what we make at full price. Today we sell 75% of what we make at full price because AI has made us more accurate in guessing what we should make.

16:46

Amazing. I was a computer teacher in 1999 and I'll never forget one of my students, for a holiday gift, gave me a book called Y2K. It's already too late. Um, and, and no, no disrespect to that former student, but I'm glad you were wrong. Yes, I want to talk about physical AI and open source, but physical AI bringing AI technology literally into the physical world that, you know, we can see and touch and interact with. And so when it comes to retail, I think of robots, I think of, I actually go back to thinking of believe it was Home Depot or maybe Lowe's. Several years back I worked on an article about how they were rolling out a robot assistant in Some of their superstores Tally to answer. Yes, exactly right. To answer questions. Kind of be a, you know, frontline triage so that the human associates could do higher level tasks and I could get my question about, you know, I need to fix a hole in my wall, I don't know how to use a hammer, what do I do? Get that triaged by the robot. Talk a little bit if you could, about kind of the state of physical AI and what retailers are, are doing. Is it robots, is it, you know, full on shopping agents that I can go into the store and they can kind of take care of it for me. What's kind of the state of play right now?

19:33

Yeah, yeah, yeah. Well, so there's a lot going on. It's all at different phases of proficiency and adoption. The like Tally type thing, which was a customer service robot in humanoid form. I would argue the AI was heavily in the humanoid form and not in necessarily the answers that Tally gave.

20:50

Right.

21:06

And so that was probably the wrong order. Like the, the bigger problem consumers faced was not knowing which screw to buy to build their deck, not talking to a human versus a website to answer that. And so I would argue the LLMs that have emerged in the last two years are a lot better at answering, although still imperfect at answering those kinds of questions. But there is a lot of mechanical automation. That's a big deal. 25% of of retail is now E commerce. Won't surprise any listeners of this podcast that a lot of the fulfillment centers are automated with robots. Right? And it used to be we had to pay kids to run around these cavernous warehouses that looked like a scene from Indiana Jones to find stuff. Now all the stuff comes to the kids, right? So the, the robots bring the shelf to the person and the, the picking job went from a, you know, walking 12 miles a day at an average shift in Amazon fulfillment center to, you know, moving good six inches from the shelf that the robot brought to you to the conveyor belt that packed it. And increasingly the kid doesn't exist at all. And the robot moves it right, right to the packing station. So these kind of automated fulfillment centers are super interesting. Amazon actually gives tours of their fulfillment center. So it's, if you want to see state of the art, you can go to these, these tours and see all those, those robots in place. But they're more visible robots to the average consumer when one of the most annoying problems in retail is taking inventory. Right. And so usually all the associates had to stay late one night a month and count everything on every Shelf. We call it cycle counting.

21:06

Right.

22:47

And it's a miserable job. Yeah. And then, and then once a year we did a huge physical inventory at least, but we also had janitors pushing floor mops around the floor. So a super cool AI automation today is the floor mops are now robots, the roombas that run around the floor cleaning the floor. And while they're cleaning the floor, they have AI computer vision cameras on them and they're taking a picture of every shelf they go by. And so they're not only cleaning the floor without the janitor pushing them up, but they're also doing that cycle count. So no one had to stay late and miss dinner, right? Yep. And so those kind of automated systems are super interesting. In club stores like Sam's Club and Costco, the highest point of dissatisfaction is the prison security check you had to go through on your way out. Right. So you, you went shopping, you bought, you know, you thought you were going for paper towels, you ended up buying a kayak. You put it in your car, you stood in line for about, you know, 37 minutes for everyone else to check out the 300 items they bought. And then after all this, you push it out the door and some scary looking security guard asked to see your receipt and you had to stand in line behind everyone else. And that security guard made sure you weren't shoplifting anything. Right. Super dissatisfying experience.

22:47

And then you get the look that's like, oh, you bought the kayak. Really? Yeah, yeah, I've been there.

24:02

Yeah. Judgy. Yeah. So today at Sam's club, none of that happens. You just push the cart out and again, a computer vision camera takes a picture of your cart and matches it to the receipt from the POS as you go out. And it's something that used to be very high friction is now friction free. And so there's, there's a lot of those kinds of physical things that, that we can see. Of course, when we go to E Commerce, we see those agents. You know, a big controversy about the future is will we still go to Amazon? Will Amazon still be the destination for shopping when the agents are in full bloom? Right. Because a lot of questions today, you don't go to Amazon and say, oh, I'm going to ask Sparky what sunscreen to buy for my vacation. Instead you might go to ChatGPT or you might go to Google and ask Gemini or ChatGPT what sunscreen for vacation. Right. And it might answer that question before you ever picked a retailer. And no surprise, all of these LLMs, these answer engines on the web, want to be retailers and they will now sell you the product that they recommend without you ever having to pick an Amazon or a Target or a Walmart. And so, you know, you're seeing those kinds of experiences more and more now.

24:06

Jason, you hit on two really salient points that for me, it's kind of like pattern recognition mapping to other industries. Right. The example of the Roomba that not only cleans the floors automatically, but does inventory as it goes. We've dived into that on previous episodes in manufacturing. Right. And talking about factory floors and optimizing the workflows and everything. But then you've got sensors that let the different automated devices, the robots, if you will, kind of perform double or triple duty. Right. So doing safety checks as they're doing assembly, that kind of thing. Just seeing that convergence of different AI related technologies coming together. And it's like this, you know, force multiplier of efficiency and cost cutting and getting things done, if you will. Right. The other thing with talking about, will you still go to Amazon Walmart, as you will, this notion of broadly speaking, is agentic AI going to, you know, do everything from kill SEO to just change the online experience to use it, you know, the way that you put it, and I've heard this for lots of other people, you know, same idea, we're not going out to the web. The web, if you will, is coming to us. Right. And the agent is bringing the things. And, and in your example, you know, I'm one of those people who will, you know, whether I have the product in mind or if I'm thinking, oh, I'm going to Hawaii, I need, you know, reef safe sunscreen for my, you know, my bald head that burns so easily. Right. I will avoid the retailer until the end because I want to find my product and then I want to find the best price and only then will I try to go in. And that, that notion of, you know, oh, well, these engine systems can do all that, so why shouldn't they also try to, you know, get the sale themselves? Yeah, it's. It spells all kinds of disruption if things head that way. For sure. Want to shift gears just a little bit, though, and ask you about open source. Open source obviously, has always been an important force in software development and the Internet and everything that we rely on all the time. You know, Nvidia recently launched Nebotron, the latest Nevotron family of models and, you know, more and more of the frontier models and the models you see on the Leaderboards are open, open source, open weights, you know, becoming open. Why is open source important to retail right now? And why are, are retailers thinking about and actually, you know, looking to use open source as opposed to just buying AI from a vendor?

25:20

Yeah, well, I think that earlier conversation we had about the pace of change is really the main reason that sort of the traditional IT model in retail is you go, oh, you have some need we didn't have yesterday. Yesterday there was no such thing as E commerce. Today there is. So let's have a shootout and let's have IBM, Oracle and SAP in here and we'll do a six month shootout and we'll pick one of these multimillion dollar pieces of software based on, which is best at this exact moment for our needs. And we'll buy it and we'll, we'll live with it and own it for the next five or 10 years. And that that model largely worked. I certainly lived through that model with many, many retailers and it built some pretty good businesses, you know, all of which have CEOs living on islands today. The challenge though is you can't have that shootout when the product at the beginning of that shootout is wildly, the outcome of that shootout when you started on Monday is wildly different than what the outcome will be on Friday because of the evolution of all of these, these things. Right. So I, I frequently walk in to a retailer and I'll give you a specific example of someone that I think is doing it really well. Walmart. Right? Doug McMillan, recently retired CEO of Walmart. Super annoying because he's, as he points out, he's younger than me and he's retired. I'm not, I'm not bitter about that at all.

27:54

One of my best friends fits that bill and more power to him.

29:11

Yeah, yeah, yeah. Doug would quickly point out, he is not my best friend. But that's a separate, separate conversation. But I used to go to meetings with him and he would say, hey, what did you, how did you use AI to prepare for this meeting? Right, Because Doug was the leader of the company, 1.6 million employees. Like we're trying to, you know, instill this, this adoption in people. And so he was trying to encourage people to embrace AI and learn about it and figure out how they can make Walmart more efficient with it. Of course, yeah. And the answer on day one, when he first asked that question in a meeting, was, Doug, we didn't, because it's illegal for us to use AI. Like it has banned all the AI tools. Right. And so that question prompted the traditional it shootout. And they go out and they pick a AI engine and they, they, they drop an AI engine. And now, okay, you all have access to ChatGPT. You're not allowed to use Google. Right?

29:14

Right.

30:05

And you're for sure not allowed to use anything that runs on AWS if you're Walmart. Right?

30:05

Right.

30:09

So then the problem becomes. Oh, no, Perplexity is actually the best LLM for commerce this week. And, uh, you know, a month ago, everyone was writing more moratoriums for, for Google because they were so far behind and they were defeated and ChatGPT had won. And four weeks later, Google is now winning every new generation. Rest empty. Yeah, exactly. And so if you just pick one, you are inevitably inferior. And so I used to go into these CEOs that are like, good news. I just did the shootout. We've totally standardized on OpenAI. And I would say that is good news. The bad news is I have more powerful AI in my pocket than you do in your whole enterprise. And they're like, what are you talking about? And I'm like, Well, I have seven LLMs in my pocket and I pay $20 a month for each of them. And every week I use a different one because they all change places about what they're good at and what they're not good at. Right.

30:09

Jason, don't take this the wrong way, but that might be why Doug's not your best friend.

30:56

One of many. One of many reasons. He also is a very smart man and a good JoJo character. So I think those are probably the, the more, more prescient reasons why he doesn't want to be my best friend. That aside, retailers today, they can't do that shootout. They can't pick that monolith. They have to be super agile and nimble, and they have to prepare for a. Capabilities are gonna come from different vendors. I joked about. Oh, and no one at Walmart wants to use aws. They're frenemies. Right. No one at Amazon wants any of these LLMs scraping their data. And so Amazon's actively blocking all of them. I am sure we will see some announcement in the near future that Amazon's partnering with a bunch of them. Right. And so all of these business decisions will change at a pace never seen before. And the poor technologists have to be able to adapt to all of that. And open source is one of the tools in the repertoire to build these product independent, IP independent solutions that are more adaptable to this new pace of change.

31:00

I'm speaking with Jason Retail Geek Goldberg. Jason is Chief Commerce Strategy Officer at Publicis Group and he's a podcaster. And as we have been discussing, he has been entrenched in all things retail, digital e commerce, mobile commerce. If it involves selling stuff, buying stuff, using technology to do it, Jason's been following it and working in it for years now. Jason, want to ask you a little bit more about agentic AI. We've been talking about this. You've been talking about this from the start. The latest Nvidia survey we mentioned, the AI and retail survey found that nearly half of retailers are either piloting or actively deploying agentic AI. You've spoken at length in our conversation about this, specifically with Rufus and Sparky. I was a little disappointed, I have to say, when you said that that third Walmart agent for the associates didn't have a cute name because I was trying to guess ahead of time, like, you know, that, that that's neither here nor there. When I think about agentic AI, whatever the use case is, my current sort of mental block and in the case of retail, is handing over the credit card. I trust an agent at this point to go out, gather information. I say agent an AI system to go out and gather information and give me recommendations and we can iterate and I find them incredibly useful for buying all kinds of stuff. But when it comes to kind of handing over the keys to the car, I'm still not sure if we're there yet. What do you see as the most promising cases, use cases for agentic AI on retail beyond what we've been discussing. Or maybe it's just extensions of what, you know, the Amazons and Walmarts of the world are already doing? But where do you see these promising use cases and in particular the ones that are really going to disrupt business first and with lasting impact?

32:02

Yeah, yeah, yeah. So watch on packing that question.

33:53

Yeah, it's a lot. I tend to do that.

33:57

I guess I would say a couple of things. First of all, we've talked about some examples of where I might use AI to assist me in making a product purchase, right. The thing to remember about products and shopping is there's a huge spectrum, right? Like when you want to pick a new car, nobody would judge you for wanting to do a lot of research before you pick that car, right? And if I go do a survey of people and say, hey, do you trust AI to pick a new your next car for you without your intervention, almost no one is going to say they trust the robot to make that highly considered purchase for them. And they want to be actively involved. Right. The other end of that spectrum is paper towels. Very few people get any joy shopping for paper towels. Right. And if you could magically wave a magic wand and say, I'm never going to run out of paper towels again, most people would just opt into that. Right. And if you said, hey, you can sign the robot up to monitor how much paper towels go in and out of my house every month and predict when I need new ones and just make sure they always show up before I need them. Most people would, would readily do that and they would have no trust issues. Right.

33:59

Not to interrupt you, but was that the idea with those buttons that Amazon sent out a while back?

35:01

Yeah. So we, we call this problem auto replenishment. Right? Like that there's a lot of products that people just don't want to run out of. When digital first became a thing, we thought, oh, man, the solution to this is this thing called subscribe and save. Right. That Amazon launched and others followed. And man, you know, I, I use the example of the, the brittle water pitcher that takes new filters, right. For 6,000 years, the way you knew to change the filter in that pitcher. I'm joking. It's not, the pictures aren't 6,000 years old, but the, the way you, you knew to change that filter is you walked by a giant stack of those filters at Costco and went, oh, shit, I've never changed the, the filter in my picture before. Right. Amazon invent subscribe and Save and were like, oh, no, Costco's hosed. Everybody's going to subscribe, get a new filter every month, and it's going to be game over. And for a year, it was like we all had to pay extortion prices to Amazon to be in this subscription program. And they started winning and sales went way down on all these, these auto replenishment consumables at brick and mortar retail. But then things regressed to the mean and sales came back and we all said, huh, I wonder what happened. What happened is we all ended up with a closet full of water filters we never put in the picture. Right? We all had subscription fatigue because the brute force of just send one every month didn't work. Based on our variable consumption, some months we use a lot more toilet paper than other months or paper towels or water or whatever.

35:06

I love bubbly water.

36:27

You and me both. Kindred spirit.

36:28

But we always are either out of it or have like three cases stacked up in the basement.

36:30

Yeah.

36:36

And it's subscribe and Save.

36:36

It's yeah, yeah, subscribe and save isn't gonna solve that problem. And so briefly, Amazon said, oh, I've got an idea. Glue a bubbly water button to your refrigerator and every time you push it, we'll bring you more bubbly water. Right. And that was called the dash button.

36:37

The dash button, right.

36:51

Yeah. I actually have some about six feet from my seat right here, so I won't get up and grab one for you, but. Fantastic. Yeah. And they short lived it turns out people don't love gluing buttons on their beautiful stain things. But also you don't need to do that. You know why? You know that Brita water pitcher we talked about earlier? It's got WI FI in it now and it orders its own. Right. And so does the dishwasher and so does the laundry machine. And so, and increasingly all of these things are happening. And oh by the way, half of us have a Alexa with a camera in it in our kitchen and it sees how often we're drinking bubbly water. So there's some near future solution when Amazon will know when you need bubbly water because of your actual consumption patterns. Right. So all of this auto replenishment is coming. There's a bunch of products you used to have to go shopping for and drive for that you're gonna, you have no trust issues. You're perfectly happy to outsource it to, to someone better. That makes it always show up. There are other products that you're not gonna trust. Day one. I use the car as the example or the, the outfit you're gonna wear on date night or the meal you're going to prepare for your girlfriend and the ingredients and all of these things or the ingredients you're going to buy for Christmas dinner, all of these things you don't trust yet. Right. And so just like no one trusted E commerce and no one wanted to type their credit card into an E commerce engine. And then no one trusted mobile and then no one trusted Facebook with their credit card. They actually still don't trust Facebook with their credit card. No cost, separate slam in those same ways. Like as these technologies evolved, people did earn trust. Right? And so the, the thing I always remind people is like at the dawn of E commerce, the best advice you'd ever get from your mom is like, don't talk to strangers on the Internet and don't get in a stranger's car. Right? And now how do we all get around? We call strangers on the Internet to get in their, their car. Like trust does come. Trust is one of the Secretly most important things in commerce. And it's going to take time for consumers to develop trust with a lot of these new methodologies, some more than others. But it is all coming, right? And so when you ask, hey, what are the big changes? The real big changes are going to come, like once you have enough trust to turn over auto replenishment. And so now suddenly the middle of that grocery store you never go to anymore, Right. Like, you're going to go buy your produce and stuff, but you're not going to get all those dry goods in the middle of the grocery store because they're just going to show up. And the marketers that used to buy super bowl ads to get you to pick the snacks, like, again, they might be trying to influence Nvidia algorithms a lot more. So maybe that's a big win for Nvidia.

36:52

Jason, did you happen to catch Joanna Stern? Ran a piece in the Wall Street Journal a few weeks back in late December about an experiment that Anthropic, the AI model company ran. They put a vending machine in the Wall Street Journal office.

39:28

It didn't do very well.

39:44

Yeah, you saw that. That's a great piece. Worked not with, but I read in the same circles as Joanna way back when when I covered the mobile phone industry. Always been a fan of her work. But it's a great piece about what went wrong and kind of hilarious and how the journalists gamed the LLM to get it to, you know, buy things that maybe wasn't supposed to. But also it kind of, I think, in lay terms kind of outlined some of these, you know, like you were talking about the technological barriers and the trust building that has to come. You know, that came for me to trust the website with my credit card, even even though I'm not yet ready to give the Agentix system the card. Let's look ahead kind of as we wrap up here. What do you see as the most critical obstacles? Technology trust. All the things we've been talking about that retailers need to overcome to maximize AI's value. Excuse me. In, you know, in the near term. And what are some of the strategies or best practices that are proving effective in addressing these challenges so far?

39:45

Yeah. Well, on the consumer side, I do think trust is the biggest challenge to adoption. Right. And so, hey, if I ask the robot for advice about the sunscreen or the car or whatever, do I believe that's better advice than I would have gotten for myself? Right. And that is going to take time. So that's a big deal. The institutional equivalent of that what's the big challenge? What's the big hurdle for the retailer? It's boring, but it's organizational change, management, right? It's, hey, you know how every big retailer in America got big through great merchants. Merchants are the people with the magic that decide what kind of products you're going to want in your kitchen next month. And they make the decisions about what to buy. And they've always been told they're the secret sauce of retail. And the good merchants are hugely successful and the bad merchants fall by the wayside. Right? And companies like the Gap became huge because Mickey Drexler, the CEO of the Gap, was the merchant prince and would, you know, be the first one to recognize that everyone wanted to wear skinny jeans or whatever the case may be. So the CEO of every big retailer in America is a merchant. Like Doug McMillan, who just retired from Walmart, was the fishing accessories merchant at Walmart when he started his career 30 years ago. And so these are merchant led organizations, right? And it's very hard to go to the CEO merchant that, you know, has a whole succession plan of fellow merchants that are going to take over for them when they retire and say, by the way, the robot's better at picking the products than you are. Right. And don't do things the way you've always done them. Don't do things the way your predecessors did them. Don't do things the way we've been doing it for 6,000 years. But do things in a totally different way is a huge, huge leap of faith. And it's not surprising that big companies with thousands or millions of employees have a lot of what I call institutional antibodies that fight these changes. Right. And so when you roll out this, this new amenity for stores, do you know what all the sales associates do? They hide it in the back room.

40:44

Sure, yeah.

42:41

Because they don't want, they don't want customers using it because then maybe they, that associate loses their job. Right? Like there's, there's all of this, this fear of change. And so, you know, the big hurdle for these companies to overcome and the one that's separating the, the best practitioners from the, the laggards is how well they is not companies that don't have it because they all have it. It's how they address it, like how they make it comfortable to take risky decisions to do things differently than they've been done. And is failure from those decisions rewarded or is it penalized? Right. Do you get fired when you do things different? Are you the, the square peg in the Round hole or are you celebrated? And so I feel like these, these leaders and these, these leadership teams that embrace organizational change management and really lean into that and say that's the fundamental problem we have to solve, do a little better. I also would say it's like a lot of things, it's a mistake to assume you have to be the first mover in all these things. Right.

42:42

That's a good point.

43:41

You know, altavista was the first mover in search and you know, not very many of US are using AltaVista today. It turns out the failure rate for fast followers is a lot lower and the success rate is a lot better. Right. Most of us would rather have Google stock than AltaVista stock. And so the magic question of your retail organization is not how do I be the first one to do all this stuff? Because it's probably not going to work for the first one. Right. The bigger question is how do I need to change my mindset so that employees would be willing to do that and what underlying, what are the precursors I need to be successful when this does happen? Right. Do I have the right data governance to even collect the data and use it with a robot? Do I have the right infrastructure? Do I have the bandwidth? Do I have the power? Like one of the big limiting factors at the moment. If every piece of commerce that could be AI powered, were AI powered, we wouldn't have enough electricity on the earth to do it. Right. And so we need, we need, you know, more efficient chips, which I think, you know, someone working on that one, and those kinds of things, you know, in order for this all to really scale. So it's not a matter of getting ready for it all happening overnight tomorrow. It's a matter of thinking about what those next smart steps are to always be prepared to be a fast forward.

43:41

So that being said, I'm going to ignore everything you just explained so eloquently and ask you to predict the future. But from the shopper side of things, from the consumer side of things, if we look, I don't know, three, four or five years out from now, what do you think will surprise, will most surprise people, consumers about how AI has changed the retail experience for them?

44:54

Yeah. There's this fundamental model we've had for an awfully long time called wholesale. Right. And wholesale is retailers buy stuff from a manufacturer. They decide what to buy, they buy it and then they store it close to the consumer and then the consumer comes and gets it. Right. And the wholesalers were the people that made the most Money in this relationship. Right. Like the Sears is the largest company in America for decades because of being a great wholesaler. And then they got replaced by Walmart and then Amazon emerged and all these, these wholesale models. I think if I follow all of these AI accelerated trends to their logical conclusion, we're not going to need wholesalers anymore. Right? Like wholesale. Like, one of the big benefits of wholesale is if you need 30 things in your kitchen, you don't want to go to 30 places to get them. So wholesalers do this thing called aggregation. They have the 30 things you need all in one place, super convenient. When the robot is buying those 30 things, the robot doesn't care that it has to buy all 30 from different places. Right. And by the way, when the robot is delivering these things, it's much more efficient, so it doesn't have to be quite as close to you to start. And so all of these things change. I think when we look back when, when, you know, the next generation looks back and goes, like, what, what did we have all wrong. Like, I think we, assuming that the model that has always dominated is going to continue to dominate, like, probably, probably doesn't play out. Like, we probably see marketplaces, we probably see social platforms being the tip of the spear for where these, these decisions happen. And I think, you know, mostly people are going to, to win, have to make something valuable that consumers want, not be one of the middlemen.

45:14

Jason Goldberg, this was a lot of fun. This was a great conversation. Information dense, but really a lot of fun. So thank you so much for taking the time to come on the pod and share what kind of, I can only imagine is just a little bit of the immense wisdom you've gathered over the years about all this stuff. For folks who want more, there's the Jason and Scott podcast. Where else can they go online? Where should they start to follow you? Follow what's going on with all things retail, e commerce, and you know what? Learn from your years of experience.

46:51

Yeah, yeah, yeah. Well, you found a number of ways to say I'm old, which I'm very grateful for. Thanks for that.

47:22

You know, as I was saying that, I was like, why am I harping on? I think I'm projecting.

47:28

Yeah, I get that a lot, especially from my family. But I am an overshare on the socials. So these days, sort of LinkedIn is my, my first point of origin. So you are welcome to follow me on, on LinkedIn or drop me a line. You know, I share a lot of these innovations there. I do go to all the industry events, and I'm lucky enough to get invited on stage a fair amount of the time for those. As you mentioned, my my much smarter co host and I have been podcasting from way before it was cool. So we We've passed our 10 year anniversary and Scott's parents spelled his name wrong. So there's only one T in Scott, which means we win at SEO, right? Just type SCOT into a search engine, you'll find our podcast. And if you type Retail Geek into any of the the socials or search engines or LLMs, you'll probably stumble across my my oversharing.

47:32

Perfect. And just to reiterate, I won't drop the URL. It's too long and it's easier for folks who want to check out the latest State of AI and Retail and CPG survey. It's on the Nvidia site. You can just type in State of AI and Retail and cpg. That'll get you faster than trying to listen to me spell out the URL.

48:19

Especially while you're driving.

48:37

You know, we one time had a guest call in while he was driving. That's a whole nother story. Jason, thanks again. This was a lot of fun. Would love to do it again. And in the meantime I look forward to looking you up on LinkedIn. Every time I want to make a.

48:38

Purchase, just drop me a line personally. I'll be your own. LLM no, I'll tell you exactly what to get.

48:50

Perfect.

48:55

Hey thanks everybody. I really enjoyed chatting with you and happy commercing everyone. Sat it.

48:56