I went to Walmart’s HQ and saw how AI is changing what people see, buy, and how fast they get it

Walmart is big, but it wants to be bigger.

The retail giant employs more than 2 million people. It serves hundreds of millions of customers each week across thousands of stores in the US and around the world. And it’s not sitting still.

Walmart is now working to speed up its business with a major assist from AI in just about every aspect of its operation.

I got to see this firsthand during its annual shareholders meeting at its headquarters in Bentonville, Arkansas, where executives and frontline workers shared how the technology is transforming retail.

AI is for everyone

For starters, Walmart is using its scale to its advantage by equipping people at all levels of the company with the tools to wrangle AI.

“We want to democratize the access to learning and democratize the access to making a difference, so that people can learn and grow,” CEO John Furner told reporters last week.

The most talked-about example was the new Code Puppy agent that Walmart Global Tech distinguished engineer Mike Pfaffenberger built and shared with the organization.

Walmart already has a dizzying number of agents and a handful of super-agents tailored for particular use cases, but Code Puppy was notable for its ability to help people across the organization vibecode their own solutions — from salaried software engineers to hourly forklift drivers.

Unlike companies that may centralize AI development within traditionally tech-focused departments, Walmart’s approach accelerates the pace at which new ideas can arise from the front lines and spread throughout the global enterprise.

“It doesn’t matter where the idea came from. It could be in Bangalore. It could be in Greater Toronto. It could be Mexico City. It could be in Wichita, Kansas. Wherever the best idea is, we should take that and scale it,” Furner said. “We just simply surface what’s already been built, and then we see the adoption rates go much faster.”

Some workers want better guardrails

Walmart’s embrace of AI is not without controversy, however.

A shareholder proposal backed by United for Respect, a coalition of retail workers, criticized the impact of AI and automation on frontline employees.

Ava Williams, an overnight stocker in Washington, presented the proposal during the Walmart shareholders meeting. She said the new AI-powered workflows push workers like her to cut corners as they race against unrealistic expectations.

“We are not asking Walmart to stop using technology. We are asking for technology that works for us, not against us,” she said.

Shareholders rejected the measure, and the company said it has multiple channels for employees to share their ideas and concerns.

The company also officially launched a credentialing program it built with OpenAI that is available to every employee, fulfilling a commitment it announced back in September.

The training program is intended to help employees build practical confidence with AI. They learn to integrate the tools into the problems they face in the real world.

The company highlighted one such case: A logistics manager who completed a Google AI certification and used those skills to develop an agent that helps Walmart identify routes that would get drivers home faster with fewer empty trucks.

Even the new delivery partnership with in-store Subway restaurants now comes with a hefty serving of AI.

Walmart order pickers already follow an AI-generated route through the store to fill the basket, and Walmart’s head of digital fulfillment, Greg Cathey, said AI now also finds the right moment in that route to queue Subway workers to make the order.

“That is the AI that’s timing everything to make sure the sandwich is going to be hot if it’s hot — always fresh,” Cathey said.

Deeper insights about what customers really want

Over at Walmart’s warehouse club, Sam’s Club Director of Consumer Insights Sue Jervis said she can now gain new details from the chain’s fast-growing member feedback community thanks to the power of multimodal AI analysis.

She expressed a visceral disgust for five-star surveys and said the conversations and video clips she gets from participants give her a much more precise understanding of what really matters to them.

Thanks to AI, Jervis can tap into the 150,000-member group and extract not just what they’re saying, but how they’re saying it — the emotional register behind their feedback about the club’s products and services.

Walmart is also learning more about customers who use the company’s new Sparky chatbot, which acts as a kind of personal shopping assistant.

“We’re learning a ton just from how they’re interacting with Sparky relative to ways that maybe historically they have interacted with us,” Chief Growth Officer Seth Dallaire told reporters.

The value of that interaction is significant enough that the company is not rushing to serve ads in the chat, focusing instead on ensuring customers are getting what they want from the experience.

“We see long natural language query strings that look very different than someone typing in ‘men’s shoes,'” he said. “That is in itself is a really interesting piece of information for us.”

Betting big on smaller things

It’s telling that on Friday, Furner handed the President’s Innovation Award to Pfaffenberger and his colleague, John Choi, for their work on Code Puppy.

“A vibe-coding tool that turns associates into engineers,” Chief Technology Officer Suresh Kumar said. “They built a tool that supports the entire company.”

Or, as Pfaffenberger asks in his documentation for the project, “Would you rather plow a field with one ox or 1,024 puppies? If you pick the ox, better slam that back button in your browser.”

Walmart, big as it is, is betting on the puppies.

Have a tip? Contact this reporter via email at dreuter@businessinsider.com or text/call/Signal at 646-768-4750. Use a personal email address, a nonwork WiFi network, and a nonwork device; here’s our guide to sharing information securely.