DoorDash handled 993 million orders last quarter. AI agents have placed "maybe several hundred" in total
SummaryOn DoorDash's third-quarter earnings call, CEO Tony Xu said outside AI shopping agents are still a tiny channel, and that when it comes to agents, "you don't want to be delusional."
· MODEL Editors · 2 min read
Some links may be affiliate links. Disclosure

The pitch for AI agents is that they will soon do your shopping for you. DoorDash just put a number on how that is going.
On the company's third-quarter earnings call on October 29, an analyst asked CEO Tony Xu about AI agents. His answer:
"My take on all things technology is you never want to fight gravity. But at the same time, you don't want to be delusional. What I mean by that is if you look at where AI agents are today, we're seeing maybe several hundred orders in total from some of these third-party agents."
Later on the same call he said again that agents "are generating several hundred orders across the time that they've been up and running. So it's been a fairly low use case." We read that in DoorDash's own transcript, which it filed with the SEC.
The other number. In the same quarter, DoorDash reported 993 million Total Orders, up 28% from a year ago (17% without its Deliveroo acquisition). That is one quarter. The "several hundred" agent orders are for the whole time those agents have been running.
Why so few? Xu said people mostly don't see outside AI assistants as the tool for an easy task like ordering food, and they still order in the DoorDash app. He also said the hard part isn't placing the order, it's making sure it actually shows up the way you expected.
The fair caveats. This is early. Ars Technica notes DoorDash's MCP connector, which lets outside AI tools order through its platform, is new and only open to a small group of beta partners, mostly for corporate lunch orders. The count also doesn't include DoorDash's own AI features, like Ask DoorDash and Text DoorDash, which it did not break out. Xu said he still sees a role for agents in bigger, "long consideration" purchases and in work done "proactively in the background," and maybe advertising inside agents someday.
So agents aren't dead. They're just, for now, a rounding error next to people tapping an app. The robot in our cartoon is a joke. The numbers are real.
Sources: DoorDash Q3 2026 results and earnings call transcript (October 29, 2026), TechCrunch and Ars Technica (October 30, 2026). Not sponsored; no affiliate links. Not investment advice.
MODEL Editors
Sources
- DoorDash Releases Third Quarter 2026 Financial Results (Oct 29, 2026)
DoorDash
- DoorDash Q3 2026 earnings call transcript, filed with the SEC under Rule 425 (Oct 29, 2026)
SEC EDGAR
- DoorDash said it's only received 'several hundred' orders from AI agents (Oct 30, 2026)
TechCrunch
- Basically no one is using AI agents for their DoorDash orders (Oct 2026)
Ars Technica
FREE · EVERY WEEK
The top AI tools this week, tested.
Every week MODEL's editors pick the AI tools worth your time, try them hands-on, and tell you what's good, what's not, and who it's for. Free, in your inbox.
Free. Unsubscribe anytime. Our picks are editorial and can't be bought.
Continue

AI News
Elon Musk: "The people deserve to have their RAM too!" (after Altman's 5% DRAM line)
Two days after OpenAI's CEO called home DRAM "quite terrible" on Cheeky Pint, Musk replied on X with a one-liner that racked up millions of views. We already covered Altman's quote. This is the sequel.
· MODEL Editors

AI News
YC told 50 AI agents to build a Minecraft civilization. They built a bureaucracy
In a YC Labs and Olam Labs experiment, 50 GPT-6.1 Sol agents got three hours and one goal. They ended up with permits, a ration ledger, and four agents spending 11 minutes appointing a "sole fertilizer operator."
· MODEL Editors

AI News
A ghostwriter's own writing got flagged as AI. Here's what detectors can and can't tell you
NPR reported on Oct 31 that a writer lost out on a book job after detectors flagged a sample she wrote herself. A June study of four popular detectors shows why one score should never decide a case.
· MODEL Editors