Inside Walmart’s AI Transformation: Empowering Work Across 1.5 Million Associates

0:04Ron Thurston:Welcome back to Voices from the Frontline AI in Action in partnership with Microsoft. I'm Ron Thurston and today we're exploring one of the largest and most ambitious AI transformations happening anywhere in retail. I'm joined by Dave Glick, Senior Vice President of Enterprise Business Services at Walmart, a leader driving a unified enterprise-scale approach to AI that empowers both corporate and frontline associates. From nanoagents and superagents to real-time translation, automation, and intelligent workflows for over 1.5 million associates, Dave gives us a rare inside look at how AI is reshaping work at the world's largest retailer.

0:52Ron Thurston:Dave, welcome to the show. I'm really happy and honored that you're here with me today.

0:58Dave Glick:Thanks, Ron, for having me. I'm looking forward to it. Pleasure.

1:01Ron Thurston:So I want to just kind of set the stage here for us. So what is Enterprise Business Services at Walmart and what's the scale and the impact that it has on all of your associates?

1:15Dave Glick:Yeah. So Enterprise Business Services is in the technology organization, but we have both technologists and engineering, as well as operations. So it's finance tech and operations, people tech and operations, global governance, customer care tech, payments, fraud, and IT. I think that's everything. And in addition to that... It's a lot of direct reports. Yeah. So that's a bunch of direct reports and a bunch of different domains, but it covers all the corporate functions. And I like to say that we ring the registers, we pay the associates, and we close the books. And so what else do you need an enterprise company?

1:53Ron Thurston:Very well said. And I think when we think about the current state of AI, and we're going to dig deep here, but thinking about how it's reshaping the retail industry and you being the largest retailer. What's your vision, say, the next two years to 10 years? What is this looking like for you? Sure.

2:15Dave Glick:In addition to the domains I support, I got the mission from Doug and Donna and Suresh to put AI in the hands of every campus associate every day. And so we've spent a lot of time figuring out how to do that. And that means letting them use the tools that are out there to do deep research, plus using AI-based tools that we've written for them as the engineering team. But the most exciting thing is that we have been doing training for all of our campus associates on the AI-based tools, and they are writing their own agent, their own agents.

2:47Ron Thurston:We're going to get to that, I think, when we talk about nano agents and super agents and what that can mean for the future. Do you see that as one of the most important long-term visions here for the brand? For sure.

3:02Dave Glick:And, you know, I've always been saying for years, putting engineers close to the customer, good things will happen. And so how close can you get an engineer to the customer? If they're the same person, that's as close as you get. And so I don't know if you've ever had this experience where you're saying, oh, we're waiting for technology to build us something, or it's below the line for engineering. That's super disempowering for our associates. And so what we've said is we have a coding agent, which you can use. And we've trained 6 ,000 associates over the last several weeks in this coding agent.

3:34Dave Glick:And we found sort of a 50 % retention rate. So we have thousands of non-engineers out there building stuff to make their lives better.

3:43Ron Thurston:I mean, the impact that that's going to have at scale for you at the enterprise level is really impressive. I can't wait to watch that happen. You're certainly leading the charge here. Yeah.

3:55Dave Glick:You know, you think about someone who spends three hours of their day validating, does this spreadsheet match this spreadsheet? Or putting together collateral for a meeting. And you can turn that from a three-hour thing to a five-minute or a 50-second operation so they can spend their time doing more interesting work. And that's very empowering for our associates.

4:15Ron Thurston:Very empowering. And that's kind of my next question was really thinking about knowing that it's such a well-documented experience about how Walmart's been so people first. How do you maintain that kind of tech-enabled work with being very people-centric? I think it's something that many brands are having conversations about today.

4:39Dave Glick:Yeah. And, you know, in our mission statement, we are people-led and tech-powered. And Doug McMillan, our CEO, has said, as long as we are serving humans, we will have humans leading the charge here. But AI is a big lever to help our humans be much more productive and much more engaged. And so my mission is to upscale everybody at Walmart. And that will help Walmart, and it will help the associates, whether they continue at Walmart or they move on to something in the future.

5:11Ron Thurston:And when you think about how to capture that feedback, what they need, what does that look like for you and all of those teams?

5:21Dave Glick:Yeah, I mean, it was interesting. So we wrote this command line coding agent. One of the engineers did it on a weekend. It's called CodePuppy. And one night we were sitting around here. The team was in town. I said, what are you guys doing tonight? They're like, we were going to go to dinner. You want a VibeCode instead with me? You're like, oh, my God, we get to VibeCode with the SVP. And so we got a case of Diet Coke and set off about 6 in the evening. We worked for about six hours. And the idea was to build the agent builder. Input an SOP, get an agent to do that, come out the other side. And it was mostly engineers, but we had a couple folks from my ops team there.

6:02Dave Glick:And they were like, we want to use CodePuppy. And so we showed them how to install it. And, you know, a week later, they all came back and said, I built this agent and I built this thing. I built this to make my life easier. And so like, I thought it was crazy because it's a command line interface. It looked like the matrix. And I didn't think the business users would like that, but they love it. the next thing that happened is I got invited to a town hall to talk about AI, of course. And I brought John, who was one of our lead live coders. And while we did the fireside chat, he built an agent in 20 minutes to show them that we were able to do that.

6:39Dave Glick:And after that, all the folks in the audience said, we want to use CodePuppy. And I started getting invitation to more town halls. And I brought John to those and they said, we want to use CodePuppy. And so, So, you know, oftentimes in engineering, we end up building something and try to get people to use it. And sort of it's a push. And it's super painful to drive adoption. This one, people are pulling, pulling, pulling. And it's sort of gotten away from us in a good way that there's more demand for the tool than there is, you know, us to teach them.

7:09Ron Thurston:Right. Exactly. Which kind of leads me to my question around, you know, maybe that's what people misunderstand about coding. But what do people misunderstand most about the frontline teams in Walmart based on your experience?

7:24Dave Glick:Yeah, I mean, we're focused on my mission is the campus associates more than the folks in the store and the warehouses. But those are frontline also. And look, people want to be able to do their jobs well. And they want to be empowered and they want to have agency. And the least empowering thing is saying, oh, you know, we were going to build something for you, but it's below the line. And so if you've got a team of five or 10 people who are doing something, you may never be above the line compared to a big ERP migration or payroll modernization or one of those things. And so by putting the tools in the hands for them, they have agency.

8:03Dave Glick:And that brings me more joy than anything else we do.

8:06Ron Thurston:Yeah. I mean, what a beautiful statement to say, we're not only here to improve your experience, but we're going to teach you a skill that you had no idea how to do yesterday. And that says something at a point in your career of like a brand new skill that may have seemed unattainable before. I really admire that.

8:28Dave Glick:It's interesting because, you know, it started out as me trying to push this across the organization or pushing from engineering. But we've seen in our people team that they've done hackathons. And I was there for judging that. And we had 42 different agents that were built over a three-week period. And many of them, very simple, like I have to validate these two spreadsheets against each other. That used to take me three hours. Now I just load them both in, and it takes me 30 seconds. And I get those stories every single day. Our finance team had a hackathon, and they gave cash prizes. And they're building, I think, what they call process agents.

9:04Dave Glick:where they're doing it themselves and they can dig into data and do a lot of research, something they've never been able to do before.

9:12Ron Thurston:And so let's jump into that deeper. So I've heard you speak about super agents and nano agents. And so for the audience, can you just share what that is? And then I've heard you speak about Sparky and how this is growing. So if you can kind of give us the basics.

9:31Dave Glick:Yeah, let's start with the super agents. We've got four super agents. One is Sparky, it's focused on customers. Marty, which is focused on suppliers. YB, which is a pun on Vibe Coding and Walmart. And finally, our associate super agent, which we'll have a name soon that we can share. I think the associate super agent is the most interesting, partly because I own it, but also because it's very diverse. You can do everything to do your job as well as manage your work life. And so if you think about the intranet, I assume you have an intranet at your company. Sometimes it's hard to find something on our intranet.

10:12Dave Glick:And so what we didn't want to do is create dozens and then hundreds and then thousands of agents that no one could ever find. And so by having the associate super agent up front, you can just go ask that whatever you need, and it will route appropriately to any of the nano agents. And so the idea is sort of a federated model, single front door, single platform, which has routing, telemetry, and sort of the undifferentiated heavy lifting. But then we're building hundreds and then thousands of these nano agents, which are domain specific, single purpose agents. And we came up with the nano agents once my friend Hari and I were sitting in the airport.

10:49Dave Glick:He's like, let's call them nano agents. And so I became the chief marketing officer for nano agents. And so we'll see if we can get it to go viral.

10:58Ron Thurston:It sounds like you could, for sure. And so to your earlier point, then you're giving kind of agency to people to build their own nano agent, correct?

11:09Dave Glick:Yeah. And so everybody has something that they spend their time doing, which they'd rather be doing more interesting work. And so if they can, and some of them are very small, like, you know, compare this spreadsheet to that spreadsheet. but another one is like and this is something we built not the users well let me take that back we built the platform and basically we loaded all of our contracts into a vector DB and now the procurement team has built dashboards and all kinds of different things saying show me show me all the contracts that are going to expire over the next six months or read the contract and see how much of it can be automatable and very interesting stuff there and that's something that we We loaded the contracts into the VectorDB, but the procurement team has gone wild on their own since we did that.

11:57Ron Thurston:And so when you think about that, Dave, the idea here of letting people build their own infrastructure of work, like how to do my job with technology. I've heard other companies maybe have some fear around giving too much power to everyone to build their own technology. So you being the biggest retailer, how do you think about that?

12:23Dave Glick:Yeah, I mean, everybody has fears. You know, one of the aha moments for me as I sat down with our head of internal audit, who obviously is very focused on this. He's like, yeah, you know, I cracked my knuckles and went and started writing some Python this weekend. And so I wrote a script, but you guys wouldn't give me an API key so I couldn't get it to production. And I was like, great, I'm glad the system works. And from that, I extrapolated to we have data governance policies. We have data privacy policies. These are all in place. We have hundreds of people whose job it is to keep us safe in InfoSec, in compliance, in audit, in legal.

13:02Dave Glick:And so let's start with those policies. You know, we spent, you know, it's been three years since the original ChatGPT moment in November of 22. We kind of spent the first couple of years trying to figure out what we could do and what would go wrong and, you know, how to be safe. And, you know, Doug got back from Davos a year ago and said, hey, Suresh, everybody else has agents. Why don't I have agents? And Suresh came to me and said, hey, we need some agents. And I said, great, I will go fast and build some agents. And the InfoSec leader, Jerry, who's in the CISO Hall of Fame, was standing next to me.

13:39Dave Glick:And he said, wait, wait, wait, wait, wait, hold on a second. And Suresh said, Jerry, you keep us safe. Dave, you go fast. And from that point, I was able to start with my most senior, my technical fellows and say, you guys start building agents. Figure out what's going to go wrong. And every single one of them that I gave this mission to came back and said, I'm doing in five days what I used to do, take me three months and a team to do. And the next week they'd come back and say, what I built last week was horrible. I did the same thing in three days and it's much better. And so the iteration speed that they were able to come up with was mind blowing.

14:17Ron Thurston:And so if you think about that, and you go back to that 2022 moment, maybe when ChatGPT was launched, what would you do differently, if anything, to where you sit today?

14:29Dave Glick:I think I would have dove in a little faster. I mean, there was a confluence of events, right? You know, I can say, oh, we should have gone faster. But the tools had a lot of problems when they first came out, right? Two, three years ago, no one ever heard of agents. No one had ever heard of MCP. No one had ever heard of A2A. We were building little chatbots and assistants that didn't work very well. And so it may be that sort of it was a nice confluence of events saying we got the mission to go fast at the same time that agentic AI had sort of sprouted out of nothing. And so, you know, I would have wanted to challenge things sooner, but it may have been that this was the right speed.

15:08Ron Thurston:Yeah, it sounds like it. Because otherwise, when you said they came back and said, I want to redo it, you may have been actually redoing it more often, more often by trying to go too fast early. But would you say now as we go into 2026, today is now is the time to move fast? For sure.

15:29Dave Glick:I mean, I think 2025 was the time to move fast. 2026 is going to be the time we scale. And I consider the last year an investment in learning. We didn't know what we were doing a year ago. Nobody knew nothing. And so by giving sort of machetes to our technical fellows and having them blaze the trail through the wilderness, we learned a lot from that and gave us the confidence to say, maybe our data governance and our data privacy is good enough. Let us go and check the edges of that and see where it breaks and then quickly respond to that rather than try to think of anything that could go wrong before we do anything.

16:07Ron Thurston:That makes a lot of sense. And I think you're right, the scale. I'm not sure everyone is exactly where you are ready to scale, but I do think there's an urgency to move quickly. And would you say, based on your expertise in the industry, is today the time to do that, to move as quickly as you can?

16:27Dave Glick:Yeah, for sure. And I talked to friends at other companies, and they're saying, we're still trying to get the basic LLMs through our security process. and I would encourage people, you gotta start trying stuff or you'll get left behind.