DoorDash Launches AI Agent for Ordering Within Apple Messages
DoorDash Launches In-App Ordering Agent for Messages
DoorDash has introduced an AI agent that allows users to place orders within the Apple Messages app. A waitlist is open for U.S. users, enabling them to order food using natural language without opening the delivery app.
Co-founder Andy Fang showcased the feature at the Dash Forward event in San Francisco. Users can send commands like "Send my usual protein bowl to the office," and the system matches the account using the phone number, retrieves past preferences and payment methods to complete the transaction. The agent can search for nearby restaurants, build shopping carts, send dish images, handle different dietary restrictions and portions in group orders, and even recommend recipes based on photos of the fridge while suggesting restocking items. The technology builds on the in-app chatbot Ask DoorDash launched in June. The company reports increased order frequency during the pilot and reminds users to check their carts before payment, as the agent may make mistakes. Bloomberg found discrepancies between quotes and the app during early testing, and the company stated that waitlisted users will receive an updated version. An Android version will be released later.
At the same time, a business-side interface was launched, allowing office assistants like Slack to place bulk orders, group deliveries, and track shipments, aimed at office catering and supply restocking. In retail, brands like Macy's have entered the app, supporting delivery in about two hours. The company also announced trials of its self-developed delivery drone, DoorDash Air, in Northern California, with initial partners including Chipotle and Popeyes, where early tests averaged delivery flights of about five minutes, with around 80% of restaurant orders fitting into the drone. The pilot for text-based ordering reportedly involves about 20,000 users.
The ordering entry has moved from the app icon to the system SMS thread, while settlement still goes through DoorDash accounts and merchant networks. Buyers include U.S. Apple users who do not want to open the app and enterprise assistants needing automated group orders, while sellers remain delivery platforms relying on commissions. The event is driven by competitive business agents, benchmarking against Uber Eats, Grubhub, as well as Meta's Muse and OpenAI's related personal agents. Beneficiaries are DoorDash, which occupies the entry point in the messaging app, and chain stores retrievable by the agent; pressured are competitors that must rely on app openings to complete conversions, and the potential for inconsistent quotes to trigger pricing regulations in Europe. European operations include Deliveroo and Wolt, where the same set of pricing errors in the U.S. is a malfunction, but in Europe could be a compliance issue.
Source: Public Information
ABAB AI Insight
After transforming its delivery network into a super app, DoorDash first introduced Ask DoorDash in the app in June, then brought the same technology into the Apple Messages thread in September, and opened a model context protocol for enterprise assistants. Andy Fang framed "not having to open the app" as the next moat: the phone number as identity, historical orders as preference models, and stored cards as settlement. The path is not to create another chatbot brand, but to occupy a system-level messaging entry, allowing search, grouping, and payment to occur within iMessage.
Capital continues to bet on order frequency. The agent reduces friction from opening the app, and the company has observed that pilot users order more frequently; the enterprise interface shifts office budgets from manual coordination to robotic ordering. Money still flows from user pre-stored payments to merchants and riders, with the platform's commission structure unchanged, while the customer acquisition interface changes. The drone pilot shifts the last mile from riders to self-developed flying vehicles, released on the same day as the text agent, indicating the company is betting on both software entry and hardware fulfillment. Retail SKUs from Macy's and others entering the scene extend the same logistics network from meals to two-hour delivery of department store goods.
This contrasts with Amazon making Alexa a shopping portal, Starbucks integrating stored cards with voice ordering, and Uber embedding trip notifications into lock screens. The food delivery industry shifts from in-app search ranking to competing for operating system messaging positions. Whoever first enables "my usual" to transact in SMS will reduce reliance on app store recommendations. The European pricing consistency issue will force agents to align menu prices before discussing natural language.
The essence is technological substitution. Menu browsing and cart clicks are replaced by natural language and historical profiles, with the app retreating from a transaction venue to a fulfillment backend. The mechanism is: once the phone number, taste, and payment are bound, opening the app becomes an unnecessary step; the cost of agent errors is wrong orders and regulation, but the benefit of migrating the entry point is a higher order frequency. The platform that occupies the messaging thread will keep competitors behind the icon.
ABAB News · Cognitive Laws
Moving the entry to SMS turns the app into a backend
"My usual" is worth more than a search bar
If quotes don't match, the agent shifts from convenience to risk