For Now, Agentic Commerce Raises More Questions Than Answers

As AI makes advances in digital commerce, payments executives are beginning to raise questions about the process and its risks. For some players, the process remains at least partly opaque on such questions as what party takes responsibility when transactions go wrong. Other pending questions include whether the industry can manage risks before common standards are established.

Such subjects have been raised at recent meetings and conferences in the industry as payments executives take advantage of such gatherings to seek out answers. Perhaps the most questions spring from the industry’s efforts to harness artificial intelligence, particularly in the form of agentic commerce. With this technology, consumers shopping online can unleash code, in the form of AI agents, to seek out particular products and then pay for them.

One big concern is how parties to the payment process recognize these transactions and how responsibility can be established when a transaction goes wrong, according to the U.S. Payments Forum, a trade group that brings together financial institutions, networks, merchants, and processors. The organization issued a meeting summary Wednesday on the unsettled matter, which it says was a major topic at its March conference in Houston.

The heart of the issue lies in the question of consumer intent, according to the group. “The next phase of agentic commerce will rely heavily on developing better ways to capture and preserve consumer intent throughout the transaction lifecycle, especially in fully autonomous or agent-to-agent environments,” the group says in its summary.

Pilots and protocols have not yet settled a number of questions, including those surrounding who takes responsibility when a transaction fails, the group says. A related issue, according to the summary, is that a set of industrywide standards regarding intent, consent, and disputes has not yet solidified. Users will “inevitably” start holding AI agents responsible for payments they did not authorize, the summary contends, requiring action on AI-specific rules regarding chargebacks.

In any case, better ways to govern agents acting on behalf of consumers is rapidly becoming a critical matter, according to the Forum. “The next phase of agentic commerce will rely heavily on developing better ways to capture and preserve consumer intent throughout the transaction life cycle, especially in fully autonomous or agent-to-agent environments,” the summary asserts.