Uber has put a number on the AI agent economy, and the number is tiny. On its third-quarter earnings call on Wednesday, the company said AI agents booked just four rides during its busiest week of October, the week of Halloween, while about 500 Uber Eats orders came through an agent in the same week. Uber handled roughly 320 million trips and orders across both services that week. Chief executive Dara Khosrowshahi called the traction “very, very early and very limited,” while insisting agents could still become a new source of demand.
## What Uber disclosed
The figures came from chief financial officer Balaji Krishnamurthy, answering the final question of the call, about building agent technology into Uber and working with outside platforms. He said AI agents booked just four rides, and that about 500 Uber Eats orders arrived the same way.
The 320 million figure covers every trip and order across Uber and Uber Eats in Halloween week, the company’s busiest of the month. Set against that volume, agent bookings are a rounding error: four rides, and a few hundred food orders. Uber did not say which assistants sent the orders, or whether any of those customers booked again.
## Why the numbers are so small
Khosrowshahi offered one blunt explanation: booking a ride in Uber’s app takes about 50 seconds, and reordering a favourite meal takes about 90. An agent that saves no time on a 50-second task has little to offer, however clever the technology behind it. The friction agents remove is barely friction at all for a ride.
Food did a little better than rides, and the gap is instructive. Ordering dinner involves choosing a restaurant, weighing options and timing delivery, which gives an agent more to do than hailing a car to a known address. Even so, 500 orders in a week of 320 million transactions barely registers, and Uber’s own verdict was unsparing.
## DoorDash said the same thing
Uber is not the first delivery platform to report numbers like these. [DoorDash](https://geroundtable.com/doordash-ai-agent-orders-hundreds/) said on its earnings call a week earlier that it had received only a few hundred orders from AI agents so far, with most customers still ordering through its app. Two of the largest marketplaces are now telling the same story: very early, very limited.
## Our Read
The numbers puncture a specific fantasy, not the whole agent story. Nobody needed an agent to book a 50-second ride, and the market has now said so twice, in hard figures. But the fantasy was never really about rides. The agent bet is on work that takes real effort, and Uber’s figures say nothing about those jobs.
What the figures do measure is the gap between agent hype and agent use, and that gap is where the money is going. Instinct raised at a $10 billion valuation, and Meta’s Muse passed 10 million users. Investors are funding a future in which [agents transact](https://geroundtable.com/visa-agentic-commerce-benchmark/); Uber’s books show the present, in which they barely do.
The agent products that are gaining ground look different from a ride-booking bot. [Perplexity’s Teammates](https://geroundtable.com/perplexity-teammates-ai-coworkers/) run continuously inside a workplace, taking on goals rather than single instructions, and similar always-on assistants are now coming from OpenAI, Anthropic and Google. The question for Uber is not whether agents book rides, but whether they eventually control the spending that surrounds a trip.
## Why It Matters
Hard numbers beat hype, and Uber has just supplied the hardest numbers the agent industry has seen. Companies planning to sell through agents, and investors pricing that future, now have a baseline: near-zero transaction volume at the world’s largest ride and delivery platform.
The stakes go beyond Uber. If agents ever handle payments and bookings at scale, the industry will need [rules for spending](https://geroundtable.com/ai-agents-approve-own-spending/), refunds and accountability that do not exist yet. Measuring what agents actually do, as Uber has now done, is the first step toward building those rules on evidence rather than on promises.
## Background
Uber disclosed the figures on its third-quarter earnings call on 4 November, a week after DoorDash reported the same pattern. The agent wave behind the numbers includes OpenAI, Anthropic, Google and Grok, all now offering assistants that act on a user’s behalf, while Khosrowshahi said about eight major assistants are on the market today, with more coming.
Source: [TechCrunch report](https://techcrunch.com/2026/11/04/almost-no-one-uses-ai-agents-to-book-their-uber-company-says/) and [The Verge report](https://www.theverge.com/transportation/1026306/uber-ai-agents-q3-earnings-disintermediation)
0 Comments