Over a few decades, consumers have moved from shopping exclusively in brick-and-mortar stores, to online, and more recently to spending billions annually through mobile apps.

Now, consumers are changing their shopping habits again. Walmart calls the new reality “adaptive retail,” which is more about personalized shopping, an expectation that retailers will anticipate what consumers want before they decide themselves, and shoppers buying specific products based on what they see influencers pitching on social media.

This evolution of consumer thinking has led Walmart to invest in newer technologies, including generative artificial intelligence. "Shopping is not just a blend of core retail and e-commerce,” says Hari Vasudev, executive vice president and chief technology officer of Walmart U.S. “It’s really a much more fluid experience.”

As Walmart sees it, online shopping has been narrowly focused. Consumers would enter a few search terms to find the exact television or jeans they were looking for. Online retailers were tasked with selling goods at a desirable price and offering efficient and speedy delivery.

But shoppers are getting more loquacious. They're asking more detailed questions about products, and they want to use new tools, including 3D that can visualize furniture in their home, for example. According to Walmart, they also increasingly want to shop with the help of virtual reality to see what clothes they want to buy look like. Meanwhile, social channels like TikTok and live streaming services including Twitch provide inspiration for shopping. "The search bar is no longer the way to shop," says Vasudev.

That’s led Walmart to pivot by announcing this month that it's starting to use predictive models and generative AI to offer a personalized version of walmart.com for every American shopper. The feature will fully roll out by the end of next year. “The way we’re thinking about applying AI and generative AI is to offer highly curated experiences and content,” says Vasudev.

The goal is to create a personalized experience for each shopper based on what they've bought previously or searched for online, matched with what's been a strong seller for Walmart or trending on social. Location and time of year would also factor in. A shopper who's previously bought goods for their dog will be shown pet costumes for Halloween. Customers in Minnesota and Arkansas will be marketed new sweaters at different points in the season, based on when winter hits.

Walmart’s generative AI efforts have thus far focused on five areas: the customer shopping experience, improving how in-store associates do their work, operational areas like supply chain and merchandising, content generation, and making software developers more productive. AI-focused tools Walmart has launched in the last few years to help employees save time have included “Ask Sam,” which lets associates ask questions like where to find a specific item in a store, and MyAssistant, a tool for corporate employees that can summarize large documents and write drafts.

In most instances, Walmart is favoring “build” over “buy” when it comes to generative AI. The company centralized machine learning capabilities in a proprietary platform, called Element, so tech staff could experiment with AI models without having to worry if they would fit into a specific vendor or cloud provider. Walmart is using a blend of in-house large language models, proprietary offerings from OpenAI’s ChatGPT and Google’s Gemini, and open-source models.

This month, Walmart also unveiled a generative AI-powered customer support assistant that relies on retail-specific large language models. The generative AI, Walmart says, will enable the chatbot to anticipate customer questions and offer more proactive solutions for handling returns, exchanges, and other problems that may arise. And in some cases, the new AI agents are proving to be twice as fast as the humans, according to Vasudev. Throughout the development process, Walmart says finding the right conversational tone was critical.

“When we have something like a chat experience, you want to make sure that the quality of the conversation feels empathetic and feels very genuine,” says Vasudev.

John Kell

Send thoughts or suggestions to CIO Intelligence here.

This story was originally featured on Fortune.com