Every agentic commerce transaction asks three separate trust questions, and the industry keeps trying to answer them with one mechanism. Discovery trust asks whether the agent surfaced the right merchant at all: an agent shopping a poisoned catalog or a spoofed storefront fails before money moves. Authorization trust asks whether the agent is acting inside the mandate its principal actually gave it: the right amount, the right merchant, the right moment. Settlement trust asks whether the payment rail underneath can carry the transaction and unwind it when something goes wrong: authorization, clearing, and a dispute path that recognizes an agent was involved.
The model matters because the layers fail independently and get fixed independently. A cryptographic mandate format strengthens authorization trust and does nothing for discovery. A verified-merchant registry strengthens discovery and says nothing about settlement. When I evaluate an agentic commerce announcement, the first question is which layer it actually addresses, and the second is which layers it quietly assumes someone else has solved.