CES 2026: Walmart, Adobe and Lenovo on selling to AI agents, not just people

Paris acknowledged the knock-on effects this is already having. “LLMs are the primary source now,” she said, noting that tools like Google’s AI overviews are pulling traffic away from publishers. “They might refer to you, but they’re not coming to you. Traffic is down across the board and it’s changing journalism.”

That reality set up the broader panel conversation with Walmart chief growth officer Seth Dallaire, Adobe Enterprise chief marketing officer Rachel Thornton and Lenovo worldwide CTO of global accounts Ed Soo Hoo. The shared conclusion was that brands are now addressing two audiences at once. Humans still respond to stories, emotion and experience. But agents interpret data structure and signals.

Thornton described it through everyday behavior. “I may have an agent buy me a Super Bowl ticket, but I go to the Super Bowl,” she said. “I may have an agent buy me shoes, but I wear the shoes.” For brands, that means thinking about human experience and agent experience at the same time. “You still need to think about what your experience is for people,” she said, “but you also have to think about what information agents need so you capture awareness and relevance when someone is looking.”

Dallaire described how that duality is already shaping Walmart’s approach. “Brands will still tell stories to humans to generate a response,” he said. “But we’re thinking about the agent in a similar way and the story may be very different.” Agents look for signals like image quality, recent reviews, price assortment and availability. “It’s still narrative-based,” he said, “but it’s a very data-driven narrative.”

The promise does not end at retrieval. “If the agent brings you to Walmart,” Dallaire said, “the expectation is that all of the brand associations you have with Walmart in the physical store will also be true in these new agent places.” That sets a high bar for fulfillment service and returns.

Commerce TV quickly emerged as the clearest test case. Soo Hoo offered a vivid example. “Think about the end of Skyfall. Daniel Craig is standing there in that pea coat and you think, ‘I want that.’ If I had a magic button, I’d click it and instantly know what it is, where to buy it and how fast it could get to me.” For him, the real prize is impulse. “Impulse is the holy grail,” he said. “Not how you push someone down a funnel.”

Thornton agreed that the magic moment is easy compared with what follows. “It’s not just capturing the impulse,” she said. “It’s fulfilling it.” Inventory ownership, delivery speed, sizing, returns and customer care all determine whether the experience holds together or falls apart.

That led to a thornier question of ownership. If you ask your TV to get you the green boots from Emily in Paris, who owns that interaction? The platform, the retailer, the brand, the viewer or the agent making the decision? The panel did not settle on one answer. “Everybody wants a bite,” Soo Hoo said, likening the moment to “a horse jumping into a pool of piranhas.” The real challenge, he argued, is making sure the ecosystem works without breaking trust.

On AI-generated ads, Sorrell pushed back on the idea that disclosure is always necessary. “We’ve been using AI tools for years,” he said, pointing to software embedded in everyday creative workflows. He acknowledged there are moments where authenticity matters more than efficiency. “There are some ads you could produce agentically,” he said, “and others you couldn’t.” Over time, he predicts economics will speed adoption. “When the economic pressure ramps up, the sensitivity will evaporate.”

The session closed with a lightning-round question about future-proofing. Soo Hoo was direct. “There’s no such thing as future-proofing. These technologies change in milliseconds.” What matters is speed of learning. “I hate the phrase ‘fail fast,’” he added. “Learn fast.”

Thornton brought it back to foundations. “Your investment in AI is only as good as your data,” she said. Customer data, taxonomy, metadata and accessibility determine whether any of this works at all. Get that right and brands have a chance to build experiences that work for humans and for the agents increasingly making decisions on their behalf.