Merchants rush to adopt AI shopping agents, Ravelin finds

Ravelin has published research showing that enterprise eCommerce merchants are rapidly adopting AI shopping agents. The study found that 44% are already integrating agentic commerce protocols.

Another 32% plan to do so within six months, suggesting the technology is moving into wider use among larger online retailers.

The findings also suggest adoption is outpacing risk planning. Only 29% of merchants surveyed said they felt very prepared for the fraud and security issues linked to agentic commerce.

Ravelin surveyed 1,504 fraud and payments professionals across 10 countries. Respondents worked for enterprise merchants with more than USD $50 million in annual revenue or more than 450 employees, and included senior management as well as fraud, risk, finance, payments, compliance, operations and product teams.

A further 42% of merchants plan to build their own custom agents. Ravelin said merchants expect agentic commerce to account for between 6% and 30% of transactions within three years.

Trust shift

One notable finding was the balance of trust between AI agents and consumers. More than half of merchants, 53%, said they trust AI shopping agents more than human shoppers, while 47% said they trust humans more.

That points to the strain fraud, refund abuse and policy exploitation are placing on retailers. At the same time, the report argues that placing greater trust in AI agents could create new weaknesses for merchants still working out how these systems should be controlled.

The risks identified include agent hijacking, fraudulent agents impersonating legitimate AI, and fake environments designed to extract sensitive data. The report also warned that criminals could deploy their own AI agents to scale fraud and abuse.

Liability questions

The research also highlighted unresolved questions over responsibility when an agent-driven transaction goes wrong. If an AI agent makes a fraudulent purchase, it is not yet clear whether liability would sit with the merchant, the agent provider, the protocol builder or the consumer.

Similar uncertainty applies to chargebacks and refunds. Merchants still face open questions about who would be liable if an agent used stolen card details or if a transaction later became the subject of a fraudulent refund request.

While merchants appear to be moving quickly, consumers are more cautious. More than half expressed concerns about fraud and security when shopping through AI agents, according to the research.

That gap between business adoption and consumer confidence could become significant for retailers deciding how quickly to roll out agent-led purchasing. It also suggests fraud controls and dispute processes may need to adapt before agentic commerce becomes routine.

Martin Sweeney, Chief Executive Officer at Ravelin, said: "Every major shift in online commerce has created new opportunities for fraud, and this will be no different.

"Merchants must take a balanced and considered approach. Fraud has consistently evolved alongside innovation. Without the right safeguards, businesses risk embedding new vulnerabilities into the next generation of eCommerce.

"The businesses that succeed will be those that move forward with a clear understanding of both the opportunity and the risk. We have done a lot of work to provide confidence to merchants that want to grow securely in the age of agentic AI commerce. As ever, managing and understanding your own data, and scaling with machine learning, will be key to preventing new fraud and growing securely."

Only 6% of merchants surveyed said they had no plans to integrate agentic AI protocols.