The gap between agentic commerce demos and actual consumer behavior got some hard numbers this week. On Uber's third-quarter earnings call, the company said that during the week of Halloween — its busiest in October — AI agents ordered exactly 4 rides across a platform that handled roughly 320 million trips and orders across Uber and Uber Eats. Agents fared somewhat better on food: about 500 Uber Eats orders were placed by an AI agent that week.
Chief financial officer Balaji Krishnamurthy gave analysts the figures, and CEO Dara Khosrowshahi was blunt about what they mean. "It is very, very early and very limited traction that we are seeing so far," he said. "Listen, we hope that there will be more traction."
The numbers land amid a wave of hype for personal AI agents that act rather than answer. Instinct, one of the buzziest startups in the category, raised at a $10 billion valuation in September. Meta's Muse became the number-one app in Apple's US App Store last month and passed 10 million users by the company's own reporting, with shopping among top activities. OpenAI, Anthropic, Google and xAI all ship agents of their own, and a crowd of smaller startups is building agents that live in text messages.
Uber is not an outlier. DoorDash said on its own earnings call last week that it has received only a few hundred orders from AI agents so far, as the overwhelming majority of customers keep using the app. The pattern across platforms is consistent: the agents are arriving, but the transactions are not.
Khosrowshahi offered a structural explanation: booking an Uber ride in the app takes about 50 seconds on average, and re-ordering a favorite meal takes about 90. An agent has to beat that — and today it mostly does not. "Every single AI system company is talking to us," he added, noting there are around eight major AI assistants on the market now and he expects that number to grow past 10.
Despite the anemic numbers, Uber is positioning itself as a willing participant rather than a gatekeeper. The company thinks agents represent "another opportunity to reach our platform" and a "net positive in terms of demand on the ecosystem," per Khosrowshahi — an approach that mirrors its historical integrations with Google Maps and other demand sources.
For the broader agent economy, the earnings-call arithmetic is a useful corrective. Consumer agents may be winning attention and valuations, but the low-friction apps people already use remain faster than any delegation flow built on top of them. Until agents can beat a 50-second booking, the agentic web stays a developer story — not yet a consumer habit.
Comments (0)
Log in to join the discussion
Log InNo comments yet