Amex wants AI agents to prove you asked for that purchase, which beats explaining why your assistant bought the wrong shoes
By Brett Gavaghan, Founder of Adaptd ·
American Express has launched an Intent Intelligence Framework that records what a customer told an AI agent to buy, and checks the purchase against it before the agent pays.
The short version
- American Express has launched the Intent Intelligence Framework to check that an AI agent's purchase matches what the customer approved.
- It's part of Amex's developer toolkit for agent-led shopping, announced on 13 October 2026, and Amex wants it to become an industry standard.
- Before letting any AI agent spend money for your business, decide its limits and require your sign-off on each purchase.
Letting an AI agent shop for you sounds handy right up until it comes home with the wrong thing. American Express is building a way to catch that before the money moves.
What is the Amex Intent Intelligence Framework?
The Intent Intelligence Framework is an American Express system that records what a customer asked an AI agent to buy and checks the purchase against it. Amex announced it on 13 October 2026, as reported by TechCrunch.
The customer's instructions, such as the merchant, the maximum budget and product preferences, are captured and signed off by the customer. The agent then pays with a secure digital credential, not the real card number. Amex's example: you ask for white Nike sneakers in size 9, the agent tries to buy Adidas, and the purchase can be stopped.
Does it matter for Australian businesses?
The framework is aimed at developers and merchants building agent shopping, so most Australian businesses won't switch anything on. Amex hasn't said where or when it reaches each market.
It's still a useful signal. Payment companies are planning for AI agents to buy things for people and businesses, and the record of what the customer approved is what fraud checks and disputes will rely on. Amex says it hasn't seen more disputes from agent purchases yet.
What should your business do about AI agents that spend money?
Any business letting an AI tool book, order or pay should set limits first: what it can buy, from whom and up to what amount.
My take: this is the payments world admitting AI agents get things wrong, and building a paper trail for when they do. Copy the idea at your own scale. Give any agent a budget, a short list of approved suppliers and a rule that it asks before paying. Nobody hands a new starter the company card on day one. Size 9 Nikes are a small mistake. A pallet of the wrong stock is a long phone call with your supplier.
Sources
Wondering what this means for your business?
That's exactly what the discovery call is for.