Guozhen AIGlobal AI field notes and model intelligence

Realtime AI News

DoorDash says AI agents have placed only 'several hundred' orders

DoorDash told analysts on its third-quarter earnings call that third-party AI agents such as Muse, Instinct and Fo have generated only "several hundred" orders in total, against roughly 3 million deliveries a day. CEO Tony Xu said customers still order through DoorDash's own app and that the company wants a healthy relationship with agent makers.

Published
DoorDash:第三方 AI 智能体累计只下了“几百单”
Image source: techcrunch.com

AI agents taking over commerce has been one of the louder narratives of the past year. DoorDash used its third-quarter earnings call to put a number on it, and the number is small: "several hundred" orders in total from third-party agents.

CEO Tony Xu, answering an analyst's question, named Muse, Instinct and Fo among the agents placing those orders. For comparison, DoorDash said in its 2025 report that it completes an average of roughly 3 million deliveries a day.

His framing was deliberately unsentimental. "You never want to fight gravity," Xu said of new technology, "but at the same time, you don't want to be delusional" — and the order count was his evidence for where things actually stand.

Xu added that users still overwhelmingly order through DoorDash's own app, but that the company wants to understand agentic use cases and build a healthy relationship with third-party agent makers. He also floated a future opportunity to advertise inside those agents.

DoorDash has been building in this direction itself. It launched a text-to-order agent in September, a command-line ordering tool in July, and the Ask DoorDash chatbot in June, which accepts prompts and photos. It has also published an MCP connector that lets other chatbots place orders, aimed at corporate accounts.

The wider landscape is split. Amazon has been blocking personal agents from its storefront, while Walmart has rolled out the welcome mat in an attempt to take business from rivals. DoorDash is positioning itself in between, keeping the door open without overstating demand.

The "several hundred" figure is a useful correction to the agentic-commerce story: the narrative has been running ahead of usage, and the default path to an order is still the native app. What to watch next is how platforms like DoorDash fold agents into their business model — through API partnerships, in-agent advertising, or both — and whether trust and checkout consistency improve enough for consumers to shift habits.

Why it matters

The number deflates the agentic-commerce narrative for delivery at least: agents are not yet a meaningful channel. For platforms the question shifts from whether to block agents to how to fold them into APIs and advertising.

DoorDashAI AgentCommerce
Back to AI Daily

Nearby Updates

All

10/30, 22:03

Jinri Yixiu launches ISHO Ring and iFlow to build an AI sleep-tech stack

Chinese sleep-tech startup Jinri Yixiu launched the ISHO Ring smart ring and the ISHO iFlow temperature-control system on October 29, alongside a personal sleep-health AI stack built on an agent app and an in-house physiological-signal model. The company says its hardware-and-software approach aims to build a leading global AI sleep-tech brand within five to ten years.

10/30, 22:30

Phase launches decentralized AI agent blockchain mainnet with $45M backing

Phase has launched the mainnet of a blockchain built for decentralized AI agents, backed by $45 million, according to Beinsure. The report discloses no investors, token design or first applications, so the launch's real weight depends on usage data still to come.

10/30, 21:00

Runware Is Building a Factory to Mass-Produce AI Data Centers

TechCrunch reports that startup Runware is building a factory to turn out prefabricated AI data centers on an assembly line, with a pitch of going online within 48 hours. The open question is whether that manufacturing model can actually deliver at scale.

10/30, 20:58

AgiBot Debuts GO-3 Preview and WITA-OMNI, Plus a Worken Unit for Robot Output

At its Innovation Day, AgiBot introduced the GO-3 Preview embodied foundation model and the WITA-OMNI omni-modal interaction model, claiming first place on three and four robotics benchmarks respectively. It also proposed Worken, a unit defined as active robots times autonomous working hours times output efficiency, to measure what physical AI actually produces.