Almost Nobody Is Letting AI Agents Order Their DoorDash Meals

The pitch is easy to picture. You tell an AI agent you want something for dinner, and it searches nearby restaurants, builds the cart, applies your usual preferences and checks out on a delivery app like DoorDash. The idea has drawn plenty of attention across the tech industry, but the latest word is that hardly anyone is actually doing it.

According to the report, AI agents play almost no role in DoorDash orders today. That gap between the buzz and everyday behavior is not surprising for a use case this new. Ordering food is already a quick habit built into apps people open on their phones, and most users see little reason to delegate a purchase they can finish in a few taps.

Agent-driven shopping also has practical hurdles. An agent has to read menus, handle customizations and substitutions, manage payment and delivery details, and make choices that users trust with their money. Each step raises questions about accuracy and control. People tend to hesitate when a mistake means a wrong order, an unexpected charge or a dinner that never shows up.

The framing of the story matters, though. Calling this "early days" points to a category that is unproven rather than dead. Delivery and grocery platforms have a clear incentive to make ordering easier, and the broader move toward AI assistants that complete tasks, not just answer questions, could make food delivery a natural testing ground. Whether that becomes routine will depend on whether agents get orders right consistently and whether people feel comfortable letting software act on their behalf.