The Future of Payments AI, Agentic Commerce and the Next Wave of Innovation

The panel opens by asking what "agentic payments" means. Panelists describe machine-to-machine commerce, where a personal AI agent searches for products and buys on a user's behalf within limits such as a spending cap. One panelist reports scanning e-commerce sites and finding that agents are often stopped by bot-protection firewalls before they ever reach checkout. The group expects friction to move away from paying and toward handling returns, wrong orders and disputes.

The middle of the discussion focuses on trust. Speakers describe signed mandates that prove what a user authorized, and a "know your agent" idea modeled on know your customer rules. They raise open questions about who absorbs losses when an agent buys the wrong item. A security researcher explains that agents read web pages differently from people, so hidden text, steganography in images and classic injection bugs can steer them. Another panelist worries that donations could be misdirected when an agent trusts unverified claims from a charity's website. Protocols such as AP2 and UCP come up, along with a call for an agent specification.

The closing points cover regulatory lag, the speed of automated fraud and the need for shared infrastructure. Panelists compare the situation to early NFC payments, suggesting one dominant player may eventually settle the fragmentation. Practical takeaways include aligning security and marketing teams on bot policy and moving fraud models from transaction-level views toward account-level views.