AI shopping agents are coming. No one is ready for them

Lack of standards around fraud prevention, security, and returns hamper growth of AI agents, experts said at the Fortune Brainstorm Tech conference.

AI shopping agents are coming. But no one is ready.

That was the consensus of a panel at Fortune Brainstorm Tech on AI shopping agents.

While plenty of people are using AI models to help discover products they might want to buy, a customer cannot easily get an AI agent to complete a purchase on their behalf due to security protocols, a lack of agentic commerce standards, and retailer policies that have sought to block third-party shopping agents, said Matt Maher, the founder and CEO of M7 Innovations, an indepedent tech research and development firm.

Melissa Bridgeford, the cofounder and CEO of Wizard Commerce, which makes AI shopping agents, said that even for product discovery, existing AI models, such as OpenAI’s ChatGPT often fall short. She noted that even when a user asks ChatGPT about a product type it might want to purchase, such as ski gloves, it responds with specific product recommendations only 9% of the time.

She said she thought OpenAI had fumbled its initial efforts to build its chatbot into a commerce platform, pivoting away from its Instant Checkout feature that allowed a user to complete a purchase directly from the chat interface. That decision also caused OpenAI’s initial retailers, such as Walmart, to pull out of the relationship.

She also said that she thought the industry was coalescing around allowing agentic commerce to proceed, but allowed that there was still no agreement over how to handle fraud, refunds, and returns—all major issues that could hold back the rollout of shopping agents.

Courtney Robinson, the head of policy and communications at open finance platform Akoya, said that liability in the case of fraud or if an AI agent undertakes a purchase that a customer claims they didn’t intend to make remains one of the biggest unsolved challenges holding back agentic commerce. “Regulation always follows innovation,” she said. “Liability is wide open right now and being negotiated company to company, but there are no standards around where liability sits when an agent buys something that maybe the user didn’t intend or ask for.”

Maher said that while many large companies will seek to protect themselves from legal liability for these kinds of AI agent errors through the terms and conditions of using their websites or agentic commerce gateways, he believed that using terms and conditions would not exempt the merchants from what he called “perceptual liability.” If his AI agent inadvertently bought a blazer from the Gap that he didn’t want the agent to buy, he said by way of example, he was likely still to complain to the Gap and expect a refund, especially if he is a loyal Gap customer.

Security is also a huge challenge. “We have a huge online fraud problem, ecommerce problem without agents and agents are only going to magnify the problem exponentially,” Norman Menz, CEO of cybersecurity company Flare, said. “The attack surface keeps expanding.” He said there were likely to be problems both with bad actors hijacking people’s legitimate agents and using them to make fraudulent purchases and bad actors spinning up their own agents using stolen identities and credit card information.

Adam Winnick, the cofounder and CEO of Finality, a company that uses blockchain technology to enable business, said he thought there would need to be new open source standards and systems around the monitoring and identity verification of AI agents and around ensuring that those agents had been empowered by their legitimate owners to conduct specific transactions on behalf of users. He said blockchain could play a role in such a solution, although he said there might be other ways to create such a system too.

The problem, many panelists said, was that the creation of such standards historically take years, while consumers are pushing to use AI agents for shopping now. “I think there is going to be a demand in the market to adopt and allow for the continued use of [AI shopping agents] before we have a solution to solve for fraud,” Menz said.

Ben Leventhal, the founder and CEO of Blackbird Labs, a blockchain-based dining rewards program for restaurants, said his company was close to being able to enable AI agents to search for restaurants and make reservations on a user’s behalf. He said in the restaurant space, he was less worried about payment fraud, because diners usually paid at the restaurant using a credit card, but that identity verification was still a key unsolved issue for AI agents.

Like Winnick, Leventhal said he thought blockchain technology could help solve this identity issue, but that other solutions might be possible and that existing identity management firms were likely to figure it out. “There is going to be an identity payload that people or their agent will carry with them,” he said.

Leventhal said that in the meantime it was likely that merchants would simply bear the risk of fraud, as they do currently in most “card not present” transactions, such as most ecommerce purchases, where no physical credit card is handed over and the customer is not physically there either.

He was also optimistic about the future of agentic commerce. “Innovators and entrepreneurs are going to find killer use cases and they are going to be impossible to resist,” he said. A lot of the current clunkiness of shopping through chatbots and using AI agents “will get abstracted away and this stuff will just become magical and any time you have magical software,” he said. “It just gets adopted.”

This story was originally featured on Fortune.com