How American Express is fixing the weak link in agentic commerce
- AI agents today can assist, recommend, and execute tasks under human-defined limits. The missing piece is a layer that makes those actions verifiable, controllable, and secure – that is what Amex’s ACE Developer Kit is built to address.
- The balance between automation, constraint, and accountability may become a key design principle of early agentic commerce.
Agentic commerce, where AI agents anticipate a user’s needs and act on their behalf, is starting to move from conceptual presentations into real-world pilots. These systems are expected to transform the entire transaction journey, from discovery and comparison to checkout and even post-purchase management. The economic upside of agentic commerce could be significant, with estimates from McKinsey & Company pointing to trillions of dollars in potential impact by the end of the decade.
But that future remains constrained by execution. That gap between decision and transaction is already shaping how these systems are designed.
It’s precisely this fault line that American Express (Amex) is targeting. With the launch of its Agentic Commerce Experiences (ACE) Developer Kit, the company is introducing a framework that enables AI agents to execute transactions on a user’s behalf but only within clearly defined, authenticated intent and control layers.
This balance between automation and constraint may become a key design principle of early agentic commerce.
Luke Gebb, EVP and Head of Global Innovation at American Express
“With the ACE Kit, the goal is to make purchases seamless without losing control or the trust, security, and service card members and Merchants expect from American Express,” says Luke Gebb, EVP and Head of Global Innovation at American Express.
“The model includes card member enrollment and authentication and gives card members the ability to manage controls directly in the Amex app – using structured intent, spend limits, merchant preferences, and tokenized credentials so the agent can only act within clearly defined boundaries set by the card member.”
At the same time, Amex is addressing the AI trust deficit more directly. Alongside ACE, it is rolling out Amex Agent Purchase Protection, extending safeguards to cover transactions carried out by AI agents, as an acknowledgment that before autonomy scales, accountability must be built in.
So, ACE is designed to do three things:
- Let AI agents transact using Amex cards
- Ensure those transactions reflect verified user intent
- Extend Amex’s protections into this new, agent-mediated layer
The real problem isn’t payment execution – it’s intent
…
