NEW YORK – At the IAB Connected Commerce Summit, Amelia Van Camp, head of agentic commerce at AI firm Mirakl, offered a timely reminder for anyone convinced AI shopping agents are already buying socks, cereal and patio furniture on humanity’s behalf: not so fast.

“So agentic commerce is in concept, in theory,” Van Camp said to Beet.TV editorial director Lisa Granatstein. “It is agents shopping on your behalf.”

Then came the reality check advertisers rarely enjoy hearing: “We are not there yet as a community, and that’s okay.”

Instead, she said the market is still in what might politely be called the figuring-it-out era. Consumers are using large language models to discover products and surface new brands, while retailers and merchants are still deciding how exactly to participate without accidentally turning their chatbot into a clearance aisle.

Discovery first, domination later

Van Camp said the louder market narrative suggests “agents are gonna take over shopping, and it’s happening now.” Her view was calmer and considerably less cinematic.

“It’s not really there yet,” she said, adding that there is still “quite a bit of legwork” before digital agents routinely shop on consumers’ behalf.

She compared the moment to early e-commerce, when everyone was experimenting, making mistakes and pretending broken checkout flows were part of the strategy.

“You see a lot of experimentation happening, a lot of errors,” she said. “So trial by error.”

Translation: the robots are learning, and apparently so is everyone else.

Glamorous future begins with your catalog spreadsheet

For brands hoping to win visibility inside AI-driven search and recommendation tools, Van Camp said the foundation is surprisingly unglamorous.

“Your catalog,” she said. “Your catalog is a huge piece of information that an AI agent, an LLM is using to determine whether to surface you inside of a conversation.”

In other words, before chasing futuristic buzzwords, brands may want to fix product titles, descriptions, attributes and inventory data. It turns out the path to the future still runs through product feeds.

She added that models also evaluate brand reputation using outside sources including YouTube, Wikipedia and Reddit, meaning marketers now have a fresh reason to care about what strangers on the internet are saying.

Van Camp said she is watching advertisers participating in OpenAI pilots for sponsored placements, calling them pioneers in the space.

Those brands, she said, are likely learning what works, what does not and how consumers react when ads show up inside conversational interfaces that were once sold as cleaner, friendlier alternatives to the ad-clogged web.

She is also studying Walmart’s Sparky commerce assistant, where advertising is already part of the experience.

Retailers want monetization, naturally

Asked how retailers should think about monetizing emerging AI channels, Van Camp said the first question is broader strategy.

“What is your plan overall for Agentic Commerce?” she said. “How are you approaching the topic?”

Large retailers with their own tools may have more room to experiment. Smaller brands, she said, should focus on discoverability, testing ads in LLM environments and improving site experiences for traffic referred by AI systems.

That means product pages, homepages and ad placements may need redesigning for a future where shoppers arrive not from search engines or social feeds, but from a chatbot that confidently recommended a blender after reading Reddit for six seconds.

Welcome to commerce’s next phase, where the robots are not in charge yet, but they are already sending traffic.

You’re watching coverage from IAB Connected Commerce Summit 2026. For more videos from this series, please visit this page.