Searching for Trust in Agentic Commerce

The episode unpacks how AI‑driven agents are reshaping commerce, demanding a new "unified trust layer" that evaluates transactions on three pillars: identity, intent and behavior. By aggregating novel agent signals, platforms can generate dynamic trust scores that replace the old deterministic risk models built around human footprints. This shift is crucial as merchants lose direct control of the checkout experience, and traditional fraud defenses—like blanket bot blocking—no longer protect against sophisticated, non‑human actors.

Liability emerges as a contentious frontier. While card networks such as Amex are beginning to assume responsibility for agent‑initiated purchases, merchants still face hard costs like incorrect shipments and soft costs tied to brand perception. The discussion highlights that liability will likely fragment across multiple scenarios—from stolen credentials to consumer dissatisfaction with AI‑chosen products—mirroring the early e‑commerce era when return rates surged. Stakeholders must therefore negotiate nuanced risk allocations that balance protection with operational feasibility.

Finally, the panel stresses the need for multi‑protocol payment orchestration. Agents will favor specific payment methods, rewarding merchants who support a breadth of networks and alternative rails. Collaborative standards development, with merchants involved from the outset, is essential to embed merchant‑first considerations into emerging protocols. By moving beyond binary accept/reject decisions toward granular trust scoring, the industry can better manage the multidimensional risk landscape of agentic commerce while preserving a seamless consumer experience.