Track · AI in Product3 min read
AI Agents Booked Just 4 Uber Rides in a Record 320 Million-Trip Week
AI agents drove just 500 Uber Eats orders and 4 rides in a 320 million-trip week, while DoorDash logged only several hundred agent orders total, both firms said on earnings calls.

Context
AI agents accounted for 4 Uber ride-hailing trips and ~500 Uber Eats orders in the last week of October.
Uber completed 320 million total trips that week, per CFO Balaji Krishnamurthy on the Q3 call.
DoorDash CEO Tony Xu reported 'several hundred orders in total' from AI agents.
DoorDash began testing a text-to-order feature bypassing its app in September.
Shoppers cite ordering errors and credit card misuse fears as reasons to limit agent use.
Four ride-hailing trips out of 320 million. That is the entire contribution of AI agents to Uber's business during the last week of October, one of its busiest weeks on record, CFO Balaji Krishnamurthy said on the company's third-quarter earnings call on Wednesday.
The number matters because it lands a hard data point on a trend product leaders keep hearing about. Agentic commerce — autonomous AI tools such as Meta's Muse and Instinct placing orders on a user's behalf — has been forecast to reshape online ordering. The platforms' own figures say otherwise, at least for now.
What did Uber actually disclose?
Agents generated roughly 500 Uber Eats orders and four ride-hailing trips during the final week of October, when Halloween demand pushed total trips to 320 million across mobility and delivery. Krishnamurthy was blunt about the gap: "It is very, very early, and very limited traction that we have seen so far."
A week earlier, DoorDash CEO Tony Xu put his company's agent-driven volume at "several hundred orders in total" — against billions of orders delivered annually. Both executives disclosed the figures on their latest earnings calls, which makes them the most authoritative usage data the industry currently has.
Why are shoppers holding back?
Business Insider's prior reporting on agent-based shopping identifies the friction points plainly:
- Ordering errors that require human correction and erode trust.
- Fears about agents "going rogue" with stored credit card details.
- Workarounds like limiting agents to price monitoring rather than checkout.
Some users do hand agents end-to-end purchasing for plane tickets and groceries. The pattern suggests agents earn autonomy on low-stakes, easily reversed transactions first — a useful segmentation signal for anyone building agentic purchase flows.
Are the platforms still investing?
Yes. Limited traction has not slowed internal tooling. DoorDash said in September it is testing a text-to-order feature that lets users order without opening the app at all — a reordering shortcut that removes interface friction rather than adding an agent layer on top of it.
Uber, for its part, is moving ahead with AI tools on its own apps despite the negligible agent order count. The strategic read for product managers: platform owners want to own the conversational entry point themselves, rather than cede it to third-party agents like Muse or Instinct that would commoditize their demand surface.
What is the takeaway for product teams?
The gap between agent hype and agent receipts is now measurable: hundreds of orders against hundreds of millions of weekly transactions, a share too small to round to a tenth of a percent. Failure modes — wrong orders, payment anxiety — are user-stated, not hypothetical. Any roadmap betting on agentic ordering in the next several quarters should plan for guardrails, reversible transactions and explicit spend limits as table stakes, not polish.
The more durable signal may be DoorDash's text-to-order test: reduce friction inside your own surface before outsourcing intent capture to someone else's agent. As agentic checkout matures, expect platforms to keep building controlled on-ramps rather than waiting for third-party agents to arrive at scale.
via feeds.businessinsider.com (Source)
More from Marcus Bennett
Show full bio
Market editor covering regulation, policy and their impact on product management at Roadmap File.
56 articles