Walmart executives see the promise of AI, but also the costs
Artificial intelligence is allowing Walmart to achieve levels of personalization it has long dreamed of. At the same time, the company is also cracking down on redundant or inefficient uses of AI internally, knowing the technology has a price tag.
AI has allowed the retailer to build experiences based on a better understanding of consumer intent and language, Walmart CEO John Furner told reporters Wednesday at the company’s headquarters in Bentonville, Arkansas, as part of its annual Associates Week and shareholders’ meeting.
“There are things we’re doing today that five years ago I don’t think we would have imagined we could do, definitely not at the speed we are, but that’s the value of just constantly thinking about, ‘Where is the customer headed?” Furner said. “It’s early, and there’s a lot that we have to learn; and we have to figure out what works and what doesn’t.”
Furner said AI is allowing employees throughout the company to more quickly complete technology projects that, in turn, could lead to more business growth.
“Every group that I meet or talk to, they tell you what they’re doing, and they tell you all the things that they are going to do in the roadmap of products they want to build and the list of things they want to improve,” Furner said. “The idea that there’s technology that can help them complete that faster is a big unlock.”
However, behind the scenes of AI is a tremendous amount of resources involved in processing large sums of data. Bloomberg reported this week that Walmart is now limiting the number of “tokens” each employee can use on an internal AI agent called Code Puppy, which can assist non-technical users with things such as spreadsheets and presentations.
Suresh Kumar, global chief technology officer and chief development officer at Walmart, told reporters that this is due to high demand from employees throughout the organization, including communications and merchandising employees and even store managers.
He said Code Puppy democratizes access to technology and allows employees to find data and get answers quickly. “Right now, we have as many … non-technical associates who are now able to do things that require technology as we have engineers,” Kumar said.
But as the use of the Code Puppy platform has skyrocketed, the company has found some of the use cases to be redundant, Kumar added. He said there are many prompts that employees keep doing again and again.
“You don’t need to keep asking Code Puppy the exact same question,” Kumar said. “We are now at a stage where not only are we able to democratize AI, but we are [also] learning enough to be able to do things more efficiently.”
The other challenge for Walmart and other retailers surrounding AI has been in managing how it shows up in outside agents such as ChatGPT. Companies like Walmart, OpenAI and Google have been trying to figure out the right moment of handoff from the AI agent to retailers. One of the biggest examples of this was the failure of ChatGPT’s Instant Checkout feature, which OpenAI abandoned in favor of letting retailers handle the checkout experience within the platform themselves.
Daniel Danker, Walmart’s evp of AI acceleration, product and design, said Walmart wanted to own the shopping functions within its purchases in ChatGPT or other agents so it can suggest complementary items and factor in other items sitting in a customer’s cart on the Walmart website that the customer also wants shipped.
“That handoff has been really important,” Danker said. “As we’ve iterated that, we’ve seen conversion go up and the experience get better.”
Still, he said he’s not concerned about losing control of consumers if they primarily end up using an agent like ChatGPT as their primary shopping engine.
“I’m far more concerned about serving customers where they are,” Danker said. “When we win with customers in the store, when we win with customers online and when we win with customers on other platforms, it’s because we won on our merits.”
Even within ChatGPT, Danker said, customers still want to make sure they are purchasing products from a reputable retailer so they can later make returns. “You want to know that if it doesn’t work out, you’ve got the wrong size, or you got too much, or it just wasn’t the right thing, that you’ll be able to return it easily. … That’s a core part of the value we deliver to customers.”
The promise of AI for Walmart is adding more levels of personalization to that value.
Danker said that personalization has previously been limited to cases like recommending a product like milk if a customer has been ordering it regularly. Now, the retailer’s AI agents can do things like identify that a customer is likely planning to make lasagna as they pick out items like ground beef and mozzarella, and recommend the other ingredients. AI can also figure out what dishes a family has in their rotation based on the items they shop, he said.
“Taking the effort out of the bits that you do every single week, and then putting some delight into the pieces that have felt like chores through true personalization, … that’s truly technology we just haven’t had before,” Danker said.
Danker said the company is seeing a shift from item-level searching to conversational shopping that is already having effects on shoppers’ engagement and behavior.
“It goes from choosing ingredients to shopping for dishes; it goes from knowing exactly how to describe the thing you need to solve your problem to just describing the problem and having solutions come your way,” Danker said. “That’s a pretty major shift in how people shop, but it’s also a pretty major shift in what people shop for.”
A different type of engagement is coming from ChatGPT customers, specifically, Danker said. Usually, prompts that lead to Walmart purchases start with prompts that don’t even involve shopping, he said. Instead of saying “What are all the things I need to buy for my camping trip?” customers may ask, “Where should I go on my camping trip?” and be recommended products along the way.
“Somewhere along that journey is an obvious commerce experience; somewhere along that journey, you’re going to need a tent, you’re going to need a certain kind of sleeping bag that’s suitable for the weather you’re likely to encounter,” Danker said. Or, he said, someone may ask ChatGPT how to get a wine spill out of their sofa. Danker said Walmart can quickly get customers anything they need to solve that problem.
“It doesn’t sound like a commerce experience, but it ends with, ‘Here’s a product that can help you,’” Danker said. “Walmart does well in those environments because we have a massive assortment that can serve almost any need.”