Amex introduces a framework to prevent AI agents from making unauthorized purchases

My take: American Express just launched its "Intent Intelligence Framework," a system that verifies the real intent behind a purchase before an AI agent completes it. The idea is simple: if you ask for white sneakers in size 9 and the agent tries to buy a different brand or size, the transaction gets blocked.

To me, this is a clear signal of where commerce is headed. According to Amex's own report, 66% of U.S. consumers say they'd use AI agents to compare prices, and 61% say they'd use them for recommendations, over the next year. Companies aren't waiting for the technology to fully mature, they're building the trust infrastructure now, for when agents routinely buy things on our behalf.

For business owners, this is a lesson in how to adopt agentic AI responsibly: efficiency means nothing without control or a way to audit what the agent actually did. And for anyone using AI for everyday tasks (restocking the pantry, comparing prices, booking something), it's worth asking what limits you'd set on that tool before handing it the keys to your card.

What controls would you put in place if you gave an AI agent permission to buy things on your behalf?