Uber says AI agents booked four rides in its busiest week of October. Its app takes about 50 seconds to book one
An agent that saves nobody time will not get used, no matter how good it is.
What happened
On its third quarter earnings call, Uber disclosed for the first time how many orders come from third party AI agents. During the week of Halloween, its busiest week in October, the company handled around 320 million trips and orders across Uber and Uber Eats. AI agents booked four rides. Uber Eats saw about 500 agent placed orders.
"It is very, very early and very limited traction that we are seeing so far," CEO Dara Khosrowshahi said. He also pointed out that booking a ride in the Uber app takes about 50 seconds on average, and reordering a favorite meal takes about 90 seconds.
Khosrowshahi still sounded open to agents. He counted around eight major AI assistants on the market, expects more than ten in time, and said "every single AI system company is talking to us." The Verge notes Uber wants to work with third parties while staying confident its customers will prefer the app.
My take
The 50 second number is the real story. When the human path is already fast, an agent has nothing to save, so people do not bother.
I use the same test when a client asks where AI agents should go in their business:
- How long does the task take a person today? Under a minute, it is probably a button or a simple trigger, not an agent.
- How many systems does it touch? Agents earn their keep on work that spans four tools and three logins, like compiling a quote, onboarding a client or reconciling payments.
- How often does it happen? Minutes saved times weekly volume is the number that justifies the build.
Ordering dinner fails all three. Preparing a weekly pipeline report from a CRM, a spreadsheet and an inbox usually passes all three.
Also worth noting for anyone selling to consumers: building an "agent ready" storefront today is a bet on future demand, not a response to current demand. Most customers still use the app.
The automations I build sit on the slow, multi-system side of that test: quote builders, onboarding flows, reconciliation, reporting and agent workflows that replace hours, not seconds. Examples are at romielwillautomate.dev.
More posts
- A 2B parameter model beat frontier LLMs at pulling fields from invoices and IDs in its makers' benchmark, at 6 to 147 times lower costNov 5, 2026
- The best AI agent finished 29 of 45 real AWS tasks. About 80% of failures came from stopping too early, not from missing knowledgeNov 5, 2026
- HubSpot's Claude connector can create and update CRM records but cannot delete them. Set its write tools to Needs Approval firstNov 5, 2026
