All posts
2 min readby Romiel Inolino

Any AI agent can now buy from Shopify's catalog, and Amex wants a signed intent before it pays. Log intent before your automations act

AI agentsecommerceShopifypaymentsautomation

Agents are becoming real shoppers. The useful lesson for builders is how both companies record what the human actually allowed.

What happened

The Verge reports that Shopify will let shoppers connect their own AI agents to their Shop account, giving more agents access to over a billion listings in Shopify's catalog, without a partnership integration. Checkout runs through Shop Pay inside the conversation, orders flow back to the Shop app, shoppers can set rules like budget guardrails or fast shipping only, and buyers must approve each purchase.

The same day, TechCrunch reported that American Express introduced an "Intent Intelligence Framework." A Create Intent API captures the customer's instructions and constraints, such as merchant, maximum budget and product preferences. The customer authenticates and signs off, which creates an Intent ID and a Proof of Intent Token. That token is used to get a payment credential the agent uses at checkout instead of the real card number. Amex gave the example of an agent told to buy white Nike sneakers that tries to buy Adidas: the transaction could be stopped. Amex said it has not yet seen more disputes from agentic commerce.

My take

Two things for operators.

If you run automations for a store, agent orders are a new channel. Check how they land in the client's CRM and reporting, since the agent, not the customer, did the browsing. Tag them so they do not vanish into "direct".

The bigger idea is the Amex pattern, and it applies far beyond payments. Before an automation does something costly, write down what the human asked for as a structured record: who, what, the limits, and when they approved it. Then check the action against that record at the moment it runs, not only at the start. A refund flow, a bulk CRM update or an ad spend change all benefit. When a client asks why something happened, you can show the approved intent next to the action. That is easier to defend than a prompt that said "be careful".

More posts