DoorDash tests AI ordering agent inside Apple Messages

DoorDash has announced an AI ordering agent that works through Apple Messages, letting customers place food orders by sending texts rather than opening the company's app.

The company said the tool accepts natural-language prompts, including shorthand references to past orders, and can respond to requests for a particular dish or a local suggestion. According to DoorDash, the agent searches nearby restaurants and assembles a suggested cart based on what the user asks for. It can also send photos of recommended food. For group orders, DoorDash says the agent can accommodate different dietary needs and quantities within a single order.

Access is not immediate. DoorDash is opening a waitlist for users in the United States to try the feature, meaning availability will initially be limited rather than broadly released.

Why DoorDash is doing this

The move positions DoorDash against Uber Eats and Grubhub, its main competitors in food delivery. It also fits a broader industry pattern in which companies are building personal AI agents intended to complete tasks without requiring users to navigate apps or websites directly. For DoorDash, a text-based agent could lower the effort involved in repeat ordering and capture requests that might otherwise go to a rival platform.

For ordinary users, the practical change is the interface: ordering becomes a conversation inside a messaging app instead of a sequence of menus and taps. Whether that is genuinely faster or more accurate than the existing app is not established by the announcement. DoorDash's description of the agent's capabilities comes from the company itself, and no independent testing or accuracy data was provided.

The drone test alongside it

DoorDash also said it will begin testing delivery drones with selected restaurants in Northern California. That is a separate initiative from the AI agent, though both point in the same direction: experimenting with delivery methods that depart from the standard courier-and-app model.

What remains uncertain

Several important questions are unanswered in what DoorDash has disclosed so far. The company did not say when the waitlist will convert into general availability, how many users will be admitted, or how the agent handles errors such as an incorrect item or an unintended order being placed. Pricing, fees, and whether the agent surfaces the same restaurant options as the standard app were also not addressed.

For restaurants, the announcement does not clarify how visibility within the agent's recommendations will be determined, a detail that matters to businesses that depend on platform placement. For creators and businesses watching the assistant trend, the DoorDash launch is a concrete example of a consumer-facing agent handling a transaction end to end, but the evidence for how well it works is currently limited to the company's own description.

DoorDash's agent is therefore best understood as a limited, waitlisted experiment at this stage, not a finished replacement for app-based ordering. The competitive logic is clear; the user experience and operational reliability are not yet demonstrated.

This report is based on DoorDash's announcement as covered by TechCrunch. Claims about the agent's capabilities are attributed to DoorDash and have not been independently verified.