How Walmart Is Reengineering AI Delivery Speed - with David Glick of Walmart

**Daniel Faggella** (0:16)
Welcome everyone to the Emerge AI in Business Podcast. Today's guest is David Glick, SVP of Enterprise Business Services at Walmart. David joins Emerge's Matthew DeMello to examine why enterprise AI efforts stall when teams are still operating on quarterly timelines. He explains how shifting to rapid real-time iteration reduces rework, shortens deployment cycles, and keeps governance aligned with actual code rather than paperwork. The conversation also shows how tasks-specific agents, automated compliance checks, and a routing layer that directs work to the right agents create measurable gains in speed, reliability, and implementation efficiency. Before we begin, a quick note for our executive listeners. Emerge invites enterprise leaders who are driving meaningful AI initiatives to share what they're learning with a peer audience. If you're moving real projects forward and want to be part of that conversation, you can learn more at go.emerge.com.expert. That's go.emerj.com.expert.
See now the conversation with David.

**Matthew DeMello** (1:20)
David, welcome to the program. It's a great pleasure having you.

**David Glick** (1:23)
It's great to be here.

**Matthew DeMello** (1:24)
Absolutely. Since we've had you on last, we've heard from a lot of Enterprise AI leaders that Enterprise AI has reached a point where technical capability is no longer the primary constraint. What limits progress instead are operating models built for slower cycles, quarterly planning, layered approvals and governance structures designed for infrequent releases. We're seeing AI systems become smaller, faster, more composable. Large organizations are being forced to rethink how speed, safety and scale co-exist inside real world enterprise environments like Walmart and at Walmart scale. Just to start things off, how are you defining stopwatch deployment metrics in a way that preserves quality, safety and reliability rather than rewarding speed alone?

**David Glick** (2:10)
Yeah, you call it the stopwatch metrics, but I've been saying we're moving from managing projects with a calendar to managing them with a stopwatch. And so what we found is with AI, AI enables rapid prototyping in minutes and hours instead of months and quarters. And we found the sweet spot for using AI. Actually, there's a lot of sweet spots for using AI, but in this particular case, the sweet spot is for prototyping. I like to post a meme, I call it the swing set meme, which is, this is what the user wanted, this is what the product manager specced, this is what the program manager thought they were on, this is what the engineers built. And you may have seen that, but oftentimes what the engineer built and what the user wanted are vastly different. And so what we found is rather than asking the user what they want, is sit with them and prototype in real time. And so we found we can able to get to the user experience in minutes and hours, rather than going back and forth. And I think a lot about, I learned a new term called OODA loops, observe, orient, decide and act. And so if we can speed up those OODA loops, saying, oh, the user would like this button moved over here. Three seconds later, it's moved over there. Or I need to add this other column, or I need to add this other workflow. And we can do that all in the same day. A lot of things that we've struggled with kind of go away. And then the second half of your question was, how do you stay safe?
You know, we have a deep security process, we call it SSP. We have data governance policies, we have data privacy policies. All of those are still in place, you know, with AI, without AI. And we are working hard to make those real time rather than something that takes months to run through all those processes. And so, you know, one of the things I believe is that if you go through these security processes, you know, they ask what databases are you using, what are the threat factors, compliance questions, all these things. And the answers to all of those are in the code. And so, rather than having a human draw an architectural diagram, or, you know, having a product manager answer all those questions, why don't we just take an agent, have it read the code, and answer those questions? And we're in the process of evolving that right now.

**Matthew DeMello** (4:23)
Very, very fascinating stuff, especially to see as you're getting there. We're seeing a lot of different deployments. You're going to take us through super agents. You talked about nano agents last time, but really, no matter, almost no matter what the AI deployment is across industries or enterprises, there's always this sense of we're leaving the status quo. We're moving from reactive to proactive or reactive to resilient. And often, I think, I think the quarterly roadmaps, or at least the quarterly timelines of corporations, are becoming kind of a stricture that is part of that status quo that's moving away. How do you envision moving away from those rigid quarterly roadmaps without eroding executive confidence or strategic alignment?