Walmart, AI, and the future of shopping, with Walmart CTO Suresh Kumar

0:00Starting a business comes with its share of ups and downs, which is why staying true to your vision is essential. a non-negotiable for Romeo and Milka Bregali, Capital One business customers and co-owners of Ra's plant-based restaurant in New York. Romeo and Milka took a leap of faith when starting their own restaurant, gutting an empty space and building it from the ground up, every pipe, every wall, every detail. But building from scratch came with a heavy financial burden, which is when they turned to their Capital One business card. With the flexibility of the card's no preset spending limit, they were able to spend more and earn more rewards while bringing their vision to life.

0:36Today, Raz's success is proof that with passion and the right support, it's possible to make your dreams a reality. Learn more at CapitalOne.com slash business cards. If you've spent any time building AI products or leading technical teams, you know this. Transformation doesn't fail because of ideas. It fails because teams can't move together. Enter Atlassian's Teamwork Collection. It has planning in JIRA, documentation in Confluence, video updates in Loom, and now AI agents in Rovo, which connects the dots across your work so nothing gets lost. It's one AI-powered teamwork platform designed for how modern teams actually build.

1:21Learn more at Atlassian.com slash TeamChanger. That's A-T-L-A-S-S-I-A-N dot com slash TeamChanger.

1:43The world's largest retailer, Walmart, has been in the news lately. Its longtime CEO, Doug McMillan, has called the company's next chapter an AI-driven transformation, deeply impacting everything from its workforce to the shopping experience. Plus, there's another big turn ahead because Walmart just announced that a new CEO will step in next year to oversee this AI evolution. And our guest today is the bridge between these eras of leadership and technology. Suresh Kumar is Walmart's chief technology officer and chief development officer. And he says AI is not an add-on for Walmart. It's core to the business.

2:25It's an integral part of how we are transforming our customer experience, but delivering on that customer experience means that we're also changing the way in which we work internally as well. And it is very, very clear that with the capabilities that these systems can provide in the workplace, the nature of work will end up changing quite dramatically. Suresh takes us inside Walmart's major push to integrate AI into every aspect of its business, online, in-store, and operations, all adding up to a completely new retail experience. This is the future of shopping. I'm Rana El-Khalyubi, and this is Pioneers of AI, a podcast taking you behind the scenes of the AI revolution.

3:23Hi, Suresh. Welcome to Pioneers of AI. I'm so excited for our conversation. Hi, Rana. Really good to be here. So we're going to spend the bulk of today's conversation talking about how you're deploying AI at Walmart. But I'd love for our listeners to get a sense of your own journey. I believe you grew up in India and you immigrated to the U.S. So did I. So I think there's a lot of commonalities in our journeys. You're absolutely right. I grew up in India. I did my undergrad work over there. I went to Indian Institute of Technology, Madras. And then after I graduated, I came directly to the U.S.

3:58to do my Ph.D. and first worked at IBM, IBM Research in New York. From his PhD at Princeton and then IBM, Suresh went on to roles at Amazon and Google. And he's been the chief technology officer at Walmart for about six years. The tech community in India has closely watched his rise. So you're kind of a business celebrity in India. You get a lot of coverage there. Do you still have strong connections? I do. We have a very large office, both in Bangalore and in Chennai. We do a ton of software development over there. And I have personal connections, too. My mom is still in Bangalore, where I grew up.

4:41I have family there. My older brother is there. My in-laws are in Chennai. So lots of reasons to go back. You know, I spent, I did an Eisenhower fellowship, and I spent about a month and a half there, like basically exploring the AI ecosystem in India. And I was just both struck but also impressed by how the country is really leaning into AI. And of course, India is one of the few countries that already has a data infrastructure. So turning on the AI layer makes a lot of sense. All right. So let's dig into all the ways Walmart is deploying AI. And I want to start with the customer-facing use cases.

5:18You guys recently announced a new partnership with OpenAI to create this, like, allowing people to shop from within chat GPT. How did this partnership come about? Oh, we are very excited about this. Look, we want to be where our customers are. And we want to make it really, really easy for our customers to shop while they're still inside ChatGPT. So what you will see very soon as a result of the partnership and the integration is that if you're in ChatGPT and, you know, you are continuing a conversation with ChatGPT, you can find all the Walmart products right there. And soon after that, you should be able to just select and complete a transaction right inside ChatGPT without leaving it.

6:08The genesis of all of this, like I said, is we want to make it really, really easy for our customers as friction-free as possible when it comes to fulfilling their shopping needs. We know that customers' habits are changing and it's no longer just about, especially if you're online, it's no longer going to be just about having a search bar where you can type in keywords and then get a list of millions of products. It's actually about fulfilling a particular need which is contextual and which is also based upon what your personal needs are as a customer. And wherever that need is felt, we want to be there.

6:50The way people are shopping is really changing, right? Right. You know, instead of, say, you know, Googling or whatever, you're now asking ChatGPT for an answer. How do you ensure that, you know, this AI bot basically recommends a Walmart product and points you in the right direction? I think it's changing search engine optimization altogether, right? It is. It's less about search optimization. It's more about understanding intent and being able to fulfill the intent. We are basically building out our own intelligent agent for shopping as well. We call it Sparky. And there, instead of having to deal with a search bar and where you type in keywords, you can have a natural conversation with Sparky.

7:41You can say that, hey, I've got six people coming over for dinner tomorrow. And Sparky will actually help you prepare for the dinner, not just the ingredients. It can understand what dietary restrictions are going to be there. Most important, it can understand what is the best way for those products to be delivered in the most seamless way to your house. Understanding what the ordering cutoff times are, what the delivery times are. So shopping is much, much more than just having a list of products at great price. That's a very important part of it. But it's also the experience. right and to understand what are the best experience that you need we need to understand walmart needs to understand what the context is the context if you're preparing dinner for six guests tomorrow is very different than you know if you are planning to go on vacation in two weeks so what is the best way to serve you depends upon what are you trying to do and what's the best way that that we can serve you for that you know i've been really thinking a lot about how companies can use AI kind of products, can use memory as a moat.

8:56Because now you're going to have, like, especially with Sparky, right, you're going to have all the history of the interactions and the conversations. How do you think about memory to optimize the shopping experience? Look, Rana, if you take a step back, the best way that we can serve our customers is to truly understand both what their preferences are and what are they trying to do. So we talked about the context, what are they trying to do? The other very important part of this is really our ability to understand who you are as a customer. And if you think about it, Walmart has more information about our customers than just about anybody else that we can use to help serve our customers better.

9:40because we serve our customers online, inside the stores, through delivery, through pickup. We also have got all the services, automotive, pharmacy, vision, our bakery. All of these things put together help us understand who our customers are at a very deep level. And that in turn allows us to be able to serve our customers in a much, much more personalized and contextual manner. And so, you know, when you talk about memory, the important thing for us is a true understanding of who the customer is and therefore what are the things that our customers really need. And we believe that this is a huge competitive advantage for us.

10:28Just about everybody is going to end up having access to the latest large language model. But what differentiates us is the fact that we serve more customers in more ways than just about anybody else. And that allows us to combine the capabilities of this larger language model with deep understanding of who our customers are, what do they want, what are their preferences, and what are they trying to do in this particular interaction, whether it happens inside ChatGPT or whether it happens on our own site or somewhere else. I love that marriage between kind of understanding, like truly understanding a customer's profile and preferences, but marrying that with intent.

11:12That's right. That's very cool. So I read a statistic that 90 % of people in the U.S. are within 10 miles of a Walmart, which is crazy, actually. So how are you using AI at Walmart's brick and mortar stores? Yeah. So first and foremost, you're absolutely right. There is a Walmart within 10 miles of 90 % of the U.S. population. Of course, we also have Walmart around the world as well. What this allows us to do is many things. One, it allows us to curate sort of the best and the highest value products that all of our customers want and place them as close to where they live as possible, right? within a Walmart super center, within a Walmart neighborhood store.

12:00So that's the first and foremost. And we are using AI to really understand what, I talked about customer preferences. So if you aggregate that, then you get a really good idea about what our customers in a particular area are really, what is it that they really want? And we can therefore curate all of our products based on that. The second thing is that having a Walmart that close to all of our customers allows us to be able to then take our supply chain, which is a supply chain that we have been building over many, many years, combining the best of grocery and perishables with our general merchandise.

12:40and now with third-party fulfillment that we do on behalf of our sellers and be able to optimize the flow of inventory all the way to the customer's homes, in many cases, through the stores itself. And what that allows us to do, again, is to deliver a seamless experience. With Walmart Express, Express Delivery, we can have products that you really need right now delivered to you in less than an hour, depending upon where you live and what your preferences are. If there is an immediate need, maybe you've run out of sugar or salt, we can have it to you before you complete making your dinner. So that is a huge advantage for us.

13:22And then the last thing that it allows us to do is to really combine online shopping and in-store into a completely seamless experience. What we call as truly omni experiences that can start online and may end up going inside the store. It might be that you start online and you don't even go inside the store because we have pickup available for you. So all of these combinations, we want to be there where you are, but we also want to provide you with the products that you want in the way that you like to have them, right? And that might change today, tomorrow, and based upon whatever context you're shopping on.

14:06Yeah. You know, I grew up around the Middle East and I suspect it's probably the same in India. My parents still to this day, if they wanted like ran out of sugar or salt, they'll just call the closest grocery store and literally within, you know, 15 minutes it's there. That's a little bit of a cultural difference in terms of how things work here. But I think there's an opportunity to create this instant kind of delivery. And it sounds like Walmart is leaning into that. Yes. So I think it's a huge desire for many of our customers that there are many products where speed becomes very, very important.

14:40But that's not the only thing, right? So there are many cases where you want much more predictable delivery. You might want to make sure that the delivery happens in a particular period of time. So we have got scheduled delivery, right? So you want to be there when the groceries get delivered. You can put it into your refrigerator as soon as it comes. So the nice thing about Walmart is that we can enable all of these things. So it's not just only quick commerce. It's not just only two-day delivery, or it's not just only about pickup. We can do everything. We're going to take a short break. Stay with us.

15:29Are you curious about the hidden side of everything? Then I have a podcast for you. I'm Stephen Dubner, host of Freakonomics Radio. Each week we hear from some of the most fascinating scholars and thinkers as we tackle big topics like how whales became the face of environmental activism, how to succeed at failing, and whether public transportation should be free. Go ahead, listen to Freakonomics Radio wherever you get your podcasts.

16:00welcome back to pioneers of ai you can also watch the show on our youtube channel back to our interview with cto of walmart suresh kumar i wanted to ask suresh about supply chain resilience when walmart delivers products all over the world how do they use ai to help the company adapt to factors like bad weather when you talk about resiliency there are several things So first of all, having a supply chain network that operates at scale allows us to plan ahead in terms of what is the optimal way to route inventory. But it also allows us to be able to react very, very quickly. So last year, when we had severe weather in parts of the East Coast, we were able to, on the fly, reconfigure our supply chain network.

16:53So that instead of delivering from one particular warehouse into a set of stores, we could reroute all of that to a different one. Same thing with our last mile delivery. If there are traffic incidents and route closures, we are able to reroute our last mile delivery dynamically and keep the customer always updated in terms of what is going on, when are they going to receive the product. But through AI, we can connect all parts of our network to every other part. So it just allows us to be able to both plan in advance and be able to react in real time. Do you also have digital twins of the stores?

17:37You do. I'd love to hear about like what technologies are you using to do that and what are the use cases? Absolutely. Yeah. So we have started building out digital twins at every level. All this optimization that I just talked about in terms of being able to plan ahead and to be able to reroute, that is possible because we have created a high fidelity model of the entire supply chain network. So we can then use that to not just plan several different scenarios and choose the most optimal one, but we can also test and see what happens if the one particular, let's say, warehouse or distribution center is closed.

18:16We can run multiple scenarios kind of immediately to say this is the most optimal rerouting to choose. Then inside what we call as our nodes. So you take a distribution center. We have rolled out a number of automated systems inside our warehouses, inside our distribution centers. And we run simulation models up front to make sure that these systems are processing inventory, whether they are receiving it, whether they are putting it on the shelf, whether they are pulling it out and creating what we call as the optimal palette to be shipped. All of those scenarios, we can test out up front. We can run simulations to figure out what is the best way to run different parts of our distribution center, as an example.

19:13And then inside the store itself, right? Inside the store, we are starting to build out digital twins so that we can deeply understand things like what is the best way for us to be able to merchandise product? What are the best facings for merchandising our product? Digital Twins also allow us to optimize the pick path for our shoppers, our Spark shoppers, or our own internal shoppers to be able to quickly go and pull product off the shelves to be able to deliver the order for you in the fastest time possible. So ultimately Digital Twins allows us to be able to quickly test out many, many different scenarios, sometimes in real time, so that we can choose the best one possible and the most optimal one possible.

20:01And when things change, we can rerun it and we can basically say, OK, now that something has changed, this is the best way for us to be able to react to it. Are you also exploring robots within either the distribution centers or the stores? I mean, there's been a lot of investments in lots of robots in stores, but also humanoid robots. And I'm curious if that's an area you're exploring? We are investing quite a bit in terms of automation inside our fulfillment centers and our distribution centers. These are purpose-built automation because these are industrial scale operations. And so there we believe that the best way to automate a facility in such a way that we remove sort of the most dangerous type of work, we can do that with custom-built automation.

20:48So So I'll give you an example. Distribution centers which handle groceries, working inside the cold chain, going into the freezer every single day for our associates used to be a very, very hard thing to do, right? So it's sub-zero temperatures over there all the time. So you had to suit up, you had to go in, hard working conditions. So there we built in automation to be able to handle product in as much of an automated way as possible. And this is very custom built, right? It is definitely AI-driven, but the hardware is custom-built. And so same thing with our stores too. We have got automated storage management systems, and these are custom-built to operate within the store environment.

21:31So that's the approach that we have been following. Given Walmart's scale, any subtle bias in any of these AI algorithms or models could affect millions of people. So I guess what concrete safeguards are in place to catch and then also correct some of these unintended consequences before they affect, you know, I don't know, pricing decisions to millions of people or whatever? This is a great question. Look, one of the cornerstones of what we are doing with our AI program is trust and safety. Whether it is trust of our customers, our own associates, by the way, our sellers, the suppliers that we work with.

22:07And so generally speaking, what we have done is to make sure that we build in evaluation, testing, safeguards at all aspects of the pipeline. Let's start with the LLMs themselves. We don't build our own foundational models. We make use of the best foundational models that are out there, whether it is by OpenAI or Gemini or by Google or the open source models or any of these things. But we don't just bring those models in-house just like that. We have built out a program where we extensively test each one of these to make sure that things like biases, hallucinations, and all of these things, they don't occur within the environment where we operate, right?

22:53So it's a very specific environment where we want these models to operate in. So first is model testing before models are let into our production. Second is around data, right? the models are only as good as the data that they operate on. So there is a lot of effort that is put in around building guardrails so that the models ingest data that has been vetted, trusted, and scanned for things like accuracy, coverage, completeness, and so on. So because we don't want models to operate on data that has been tainted or is incomplete, because then you can't trust the answers that it gives out. And then the last part of it is on the output itself.

23:42So we have human in the loop where we take, you know, a statistical sample during operational time and during build time, we take a significant amount of the output that comes out, starting with all the output when we start out something brand new, pass it through human evaluation to see whether the combination of data and models is actually producing the right results or not. And then we go back and tune. And once we get it right, we still have a small fraction of the outputs that these models generate to always go in front of a human evaluator because these models can drift, the data can drift over a period of time.

24:22So that's how we make sure that across the entire pipeline, data, model, output, that we have safeguards everywhere. Coming up, more on how Walmart sees its AI edge in a steeply competitive retail space. Stay with us.

24:56meet nicole nicholas capital one business customer and co-owner of anset uncles a plant-based restaurant and community space in brooklyn new york that got its start from a need for unity. The inspiration, it was born from the desire to create a space that felt like home, where we can connect community culture, good food, and come together with family and friends. That's how we birthed aunts and uncles. Nicole and her husband, Mike, were fulfilling their dream of bringing people together out of their home kitchen. But they soon learned that the demand for community was greater than they knew. It became overwhelming and we were like, we need home, but not in our actual home.

25:32We realized that there was also a need in our community for something bigger in our neighborhood. So we had to find a place. Moving from a home operation into a storefront was a huge next step, but Nicole and Mike were able to take it on with the help of Capital One Business. It's not for the weak. As a small business, finding resources is super important because that's the way you'll be able to manage and scale. We would have never done that without having Capital One to be able to help us along the way. The cashback rewards are very helpful. You know, it just gave us that runway to be able to breathe a little bit.

26:06Then you get to focus on the cooking of the food and making the experience great. To learn more, go to CapitalOne.com slash business cards.

26:19So Amazon is still the leader when it comes to e-commerce, but Walmart is showing impressive growth in this space. What would you say is Walmart's competitive advantage? There are two bigger advantages. I think the first advantage clearly is that customers care about having access to great quality products. They care about trust, they care about convenience, and they care about the experience overall. So that is one big advantage that Walmart has because we serve customers no matter how they like to shop and how they want their products, right? Whether it's online, offline, inside the store, delivery, pickup, all of these things.

27:00So that's a huge advantage. The other big advantage that we have is that serving our customers in all of these different ways means that we have a much deeper understanding of what our customer needs are, what their preferences are. And I talked about the contextual piece. Translating that context into action requires not just an online presence, but it also requires us to have the inventory in the right place, which is where are thousands of stores having a Walmart within 10 miles of 90 % of the U.S. population. That's where the advantage comes in. So ultimately, it is not about e-commerce or stores.

27:42It's about how you deliver the entire experience. Kind of marrying the physical with the digital and making that a seamless experience. AI is a vital part of Walmart's future, and Suresh is the person shaping that for the global giant with$680 billion in annual revenue. After our interview, longtime Walmart CEO Doug McMillan announced he'll retire from the company in January. Just a few weeks before that, McMillan made a major announcement about AI and Walmart's workforce at the company's Arkansas headquarters. He said he couldn't imagine any job that wouldn't be impacted by AI. And that Walmart's plans to grow as a company do not necessarily include adding more people to its workforce.