DoorDash AI Agent Takes Food Orders by Text Message

DoorDash wants customers to skip its app entirely. On Wednesday, the delivery company announced a new AI agent that takes food orders by text message. It runs inside Apple Messages instead of DoorDash's own interface. For now, people in the U.S. can sign up for a waitlist to try it.

How the agent works

You write what you want in plain language, and the agent turns it into an order. DoorDash's own example is a message like "order my usual." The company says the agent can work out that this means the order a customer normally places on a Friday night, and then act on it.

The agent also handles new orders. Users can name a specific dish or ask for a recommendation nearby. The agent then searches local restaurants and builds a suggested cart based on the request. It can also text photos of the food it recommends, so users can see the dishes before they commit.

Group orders are part of the pitch too. DoorDash says the agent can combine different dietary preferences and different quantities in a single order. Anyone who has collected lunch orders across a team chat knows this is usually where ordering gets messy. That makes it a sensible use case to highlight.

Competition and the push toward personal agents

The launch has a clear competitive angle. DoorDash competes directly with Uber Eats and Grubhub, and an AI ordering agent gives it a way to stand apart. The bet is that convenience at the very first step, placing the order, can count as much as selection or price.

The move also fits a broader shift in the industry. More companies are pushing personal AI agents that complete tasks for users, so people no longer have to move through apps or websites themselves. The same trend shows up in efforts like OpenAI's always-on ChatGPT agents. A messaging thread is a natural place for this kind of tool. Many people already spend much of their day there, and there is no new interface to learn.

Drones on the side

DoorDash paired the agent with a second update. The company will begin testing its delivery drones with select restaurants in Northern California. It did not formally connect the two projects. Both, however, target the same thing: less friction between wanting food and getting it.

Our Take

DoorDash's agent is a small, practical example of where consumer AI agents are heading. It is not a general assistant that tries to do everything. It is a narrow tool built around one frequent task, and that narrow scope may be its main strength. Ordering dinner is low-stakes compared with moving money or handling sensitive files. That makes it a reasonable place for consumers to get comfortable handing tasks to an agent.

The feature still depends on the agent reading intent correctly. "My usual" only works if the system remembers past behavior, and that raises familiar questions about how much personal data an agent should keep and how it uses it. Recent controversies such as the dispute over Meta's Muse agent show how quickly trust can become the central issue for agents that sit close to private conversations.

A few things are worth watching from here. Uber Eats and Grubhub may respond with similar tools. It is also unclear how the agent will handle wrong or misunderstood orders once real users test it at scale. Finally, DoorDash could expand the feature beyond Apple Messages once the waitlist phase ends.