Based on American Express’s press release. Interview remarks and survey figures given to TechCrunch are labelled REPORTED.
American Express has published a framework called Intent Intelligence. It sets out how to capture and confirm what a customer wants before an AI agent buys something for them, and then keep checking against that through the rest of the transaction [1].
Amex says it is making the framework public to “help advance a common industry approach for intent in agentic commerce” [1]. “Agentic commerce” means AI agents shopping and paying on a person’s behalf.
The problem it is trying to solve
AI agents can make mistakes. TechCrunch notes that large language models do not always behave predictably, so an agent could buy the wrong thing or spend more than it should [2].
Amex’s idea is to record the customer’s instructions first, have the customer sign them off, and tie the payment to those instructions.
How it works
Amex describes four steps [1]:
- Create and confirm intent. A Create Intent API records the customer’s instructions and limits, “such as Merchant, maximum budget, and other preferences”. The customer then proves who they are and signs off on what the agent captured.
- Proof of intent. The system creates an Intent ID, a unique reference for that request, and a Proof of Intent Token that holds its main limits.
- A limited payment credential. The token is used to get a secure digital payment credential. The agent pays with this “instead of the customer’s actual Card Number”, and the credential is limited to what the customer approved.
- Checks all the way through. The Intent ID and token can be used later to check that a purchase matches what was authorised, helping with “fraud prevention, authorization decisioning, and dispute resolution”.
Amex says the difference from other approaches is that the card issuer sees the approved intent earlier, “before an agent executes a purchase” [1].
“Building trust into AI-powered transactions is critical to the adoption of agentic commerce,” said Luke Gebb, Head of Global Innovation at American Express [1].
The framework is part of Amex’s Agentic Commerce Experiences (ACE) Developer Kit. Amex also says it is working with EMVCo, the card industry’s standards body, and supporting the x402 Foundation’s work on internet payments standards [1].
What Amex told TechCrunch
- Hannah Bozian, Amex’s VP of Agentic Partnerships and Strategy, gave an example: if a customer asks for white Nike sneakers in size 9 and the agent tries to buy Adidas shoes, Amex could stop the transaction [2].
- Amex said it has not yet seen an increase in disputes caused by agentic commerce, but wanted to act early [2].
- In Amex’s Trendex survey, 66% of US consumers questioned said they were likely to use AI agents to research the best price, and 61% to get a product or service recommendation, over the next 12 months, the company told TechCrunch [2]. TSN has not seen the survey itself.
What this means
- Your instructions become the limit. If it works as described, an agent can only spend within what you signed off.
- The agent never sees your card number. It uses a limited credential instead.
- It is a proposal, not yet a standard. Amex hopes others adopt it; that has not happened yet.
Sources
- American Express via Business Wire, “American Express Publishes Intent Intelligence Framework for Trusted Agentic Commerce”, 13 October 2026. https://www.businesswire.com/news/home/20261013636634/en/
- TechCrunch (Sarah Perez), “Amex introduces a framework to prevent AI agents from making unauthorized purchases”, 13 October 2026. https://techcrunch.com/2026/10/13/amex-introduces-a-framework-to-prevent-ai-agents-from-making-unauthorized-purchases/
