On September 30, 2026, DoorDash launched a text-to-order AI agent inside Apple Messages. A user can send a prompt like "order my usual," ask for a specific dish, or request a local recommendation. The agent searches local spots and suggests a cart from the prompt. It can text photos of the food it recommends and, for group orders, handle mixed dietary preferences and different quantities in one order. DoorDash opened a waitlist for U.S. users to try the feature. The agent takes over the two places where a brand normally wins or loses: the menu a customer scrolls and the checkout flow a customer completes. It builds the cart from the product data it can read, so the brands that land in the cart are the ones that describe what they sell in a form the agent understands.
Mastercard added an AI-transaction probability score and a fraud intelligence layer to Agent Pay in September 2026, with Cloudflare and Skyfire as partners. The score is a trust signal for the transaction itself. When an agent initiates a purchase, the payment layer decides whether to approve it.
What should you test? Start with the data an agent would need to assemble a cart from your catalog. That means product names, descriptions, prices, availability, and order minimums, all in a machine-readable format. Then test the payment handoff. Does your order data pass a fraud check when an agent submits it? Can an agent complete a purchase without a human stepping in? The person who owns revenue should own these tests.
DoorDash's launch and Mastercard's update sit on the same line. An agent now assembles the order, and a payment network scores it, before a person reviews either one.