Uber: 4 rides and 500 Eats orders came via AI agents in a 320M-trip week

Very few customers are using outside AI agents to buy from Uber or DoorDash, according to what both companies told investors in the past week. TechCrunch reported on 4 November 2026 that on Uber's third-quarter earnings call, CFO Balaji Krishnamurthy said only four rides were ordered through an AI agent during the week of Halloween, Uber's busiest week of October.

The numbers

• Uber, per TechCrunch: about 320 million trips and orders across Uber and Uber Eats in Halloween week; four rides and 500 Uber Eats orders placed by an AI agent.
• Uber CEO Dara Khosrowshahi, quoted by TechCrunch: "It is very, very early and very limited traction that we are seeing so far."
• DoorDash, per Ars Technica on 30 October: CEO Tony Xu said "AI agents are generating several hundred orders" across the time they have been running, on just under 1 billion orders in the third quarter of 2026. He called it a "very low use case".
• Uber's share, worked out here from the reported figures: 504 agent-placed orders out of roughly 320 million is about 0.00016%. That arithmetic is mine, not Uber's.

What the numbers leave out

Ars Technica notes that DoorDash launched its Model Context Protocol (MCP) connector, a standard way for outside AI tools to place orders, less than a month before the call. It is still limited to a small number of early beta partners and is mainly used for corporate lunch orders. The figure also excludes DoorDash's own assistant, Ask DoorDash. DoorDash has said Ask DoorDash baskets get built "five times faster" and average nearly 50% higher basket value, though that is the company's own claim and it has not published usage totals.

Neither source says how either company identifies an agent-placed order, or whether agents driving the apps through a web interface are counted. Those definitions are unknown here, so treat the totals as lower bounds on agent activity, not exact counts.

The executives' reasoning

Khosrowshahi pointed out, per TechCrunch, that the average time to book a ride in Uber's app is 50 seconds and re-ordering a favourite food takes 90 seconds, so an agent may not save anyone time. He still called agents "another opportunity to reach our platform" and expects a net positive for demand. Xu told analysts that for a task as simple as ordering food, customers "tend to see that come directly to DoorDash", and saw more promise in long-consideration purchases where an agent works in the background.

What a team can take from it

• Short, frequent tasks that a good app already handles in under two minutes are a weak case for an agent channel. Longer, research-heavy purchases are where the two CEOs expect agents to matter.
• Count agent-originated transactions separately, as these two companies now do, so you can see whether an MCP or API investment is being used.
• Check how new the integration is before reading a low number as demand. DoorDash's had been live for under a month.