The world’s largest retailer is collapsing software cycles with AI, shrinking year-long tasks to days. Curiosity and iteration now drive speed, making tech no longer the bottleneck.
“Running projects with a stopwatch instead of a calendar.”
That is how David Glick, Walmart’s SVP of Enterprise Business Services, describes the company’s shift to AI.
What once took teams of engineers a year can now be produced in days or even hours, and the effect is already changing how Walmart designs, tests, and delivers software across the enterprise.
There was a moment, he recalls, when the engineers claimed the code was nearly ready, fifteen bugs left, almost clean, while the product team came in waving their arms that nothing worked. The old pattern would have meant another cycle of meetings, test plans drafted over months, and perhaps a year to build a framework that would catch the failures.
Instead Glick called an engineer on a Friday afternoon at five, pointed AI at the codebase, and by eight they had an automated test plan running.
“Can you imagine building a whole test plan and test framework that might take a whole team of six to eight engineers a year to do this? And you know, a couple engineers sort of in their off hours were able to do this in a few weeks. It’s kind of mind boggling.”
A task that once absorbed a team for a year collapses into the off-hours of a weekend. The system may be rough, the first output flawed, but in the rhythm of stopwatch time, roughness is tolerated because iteration is constant.
The slow, almost ritualistic phases of pre-discovery, discovery, UX, testing were already famous for producing the wrong thing after six months of labor. What if that same wrong thing could be produced in three hours?
“Instead of having three months of product discovery, we go build wireframes and a clickable prototype in three hours or in three days, and then we can show it to the business, and they’ll say, we don’t like it. Now we know six months earlier that you don’t like it,” he told a group of us gathered at the AI Infra Summit.
The future, he kept insisting, belongs to the curious rather than to those with the most senior titles, or even to those with the deepest coding experience. It’s the the ones willing to press prompts, run tests, and discard their own outputs until something holds.
What we found is that people, whether in a product team or the design team or the engineering team or the business team, the people who are succeeding most with AI are the ones who are most curious.
In Glick’s experience, things are working when an engineer no longer waits for requirements to descend from the product team. A designer can prompt for code, a business user can generate their own wireframe. The edges of responsibility blur, sometimes uncomfortably, but the net effect is speed.
“We invite the business users and say, Come watch what we’re doing. You can tell us what to do, and we’ll do it in real time there. And if you can get a business user to sit with you for three or four or five hours, whether it’s at night during the day, they’ll be a believer.”
Glick turns the conversation to something unexpected by talking about the weight of governance in the midst of the broader innovation push toward AI, noting that it was often the reason they expected not to be able to move and iterate so fast.
Walmart can’t risk leaks or compliance failures but even here teams have found a way to adapt. “We called the Chief Compliance Officer and said, ‘Hey, we need you know, if we can build an agent in a week, we can’t take you two months to approve our agent.’ And so she went in and, built a new process, and still equally as safe, it leveraged AI, leveraged technology equally as safe, but now the backlog is zero days instead of 60 days.”
This is what it means when Glick says the company is running projects on a stopwatch. Every artifact of corporate technology, from test plans, security reviews, compliance sign-offs, wireframes, prototypes, shrinks from months to days.
It doesn’t mean the work is easier per se but it does mean the cycle of bad outputs and discarded drafts can spin so much faster that the third or fourth iteration actually arrives before interest is lost.
For the associates outside engineering, the change is equally radical. They are not asked to become expert coders. They are asked to use AI like a second language. Drafting memos, generating analyses, automating the small irritations of their jobs. And interestingly, the qualities Glick names are not technical. “Curiosity, persistence, resilience, and grit.”
In terms of the path ahead, “I can’t even think about the next one to three months,” Glick said. But when pressed he sketches an outline where code itself becomes an even smaller part of the lifecycle, testing and security reviews automate themselves, product requirements are generated and iterated in hours.
We are only limited by our creativity. We’re no longer limited by the number of servers we can get, the number of engineers we can hire where we can do amazing things, which we never thought of.
Glick described what it all looks like on the ground when it’s executed well. An engineer with a laptop walks into a Walmart store or fulfillment center. They spend a few hours with an associate, watch how the work is actually done, identify the friction points, and build an agent on the spot.
What slows this down now is not infrastructure, he says. Servers are provisioned instantly, code is generated on demand. The bottleneck is human.
“Actually writing code… we know how to do that, and it’s getting easier and easier using AI, but it is, in fact, a change management [problem]… we’re moving people’s cheese. And so how do you push this along?”
At this scale, velocity itself becomes a kind of infrastructure.
The company that once mandated barcodes now mandates curiosity. The bottleneck is not servers or engineers but the willingness to imagine. And the calendar, the months, the quarters, the years that used to govern corporate technology, is suddenly obsolete. The watch is running.
“Everybody at Walmart, certainly the folks at Walmart corporate, should be using AI every day, and they’re building agents to make their job better.”
