Uber says only four rides were booked by AI agents in its busiest week

Uber has put a hard number on how often people actually let AI agents book rides for them, and it is strikingly small. According to TechCrunch, the company said on its third-quarter earnings call this week that just four rides were ordered through an AI agent during its busiest week in October, the week of Halloween.

The delivery side fared somewhat better but remained marginal. Chief financial officer Balaji Krishnamurthy told analysts that Uber processed roughly 320 million trips and orders across Uber and Uber Eats that week, and only 500 of the Uber Eats orders were placed by an agent, TechCrunch reports.

CEO Dara Khosrowshahi characterized the uptake as "very, very early and very limited traction," while adding that the company hopes usage grows.

The math is unforgiving

Four rides out of a week that generated hundreds of millions of transactions is not a nascent trend; it is a rounding error. Even the 500 agent-placed food orders, ten times the ride count, disappear into a weekly volume of that size. If agentic commerce were meaningfully arriving, a platform with Uber's scale and integration appetite would presumably be among the first to see it.

Apps may simply be too fast

Khosrowshahi offered a plausible explanation for the apathy, as TechCrunch notes: the average Uber ride takes about 50 seconds to book in the company's app, and re-ordering a favorite meal takes around 90 seconds. When a task is already that quick, an intermediary agent has almost no time to save, and adds a layer of uncertainty about whether the order came out right.

DoorDash sees the same pattern

Uber is not an outlier. DoorDash told investors on its own earnings call last week that it has received only a few hundred orders from AI agents in total, with the overwhelming majority of customers still ordering directly through its app, according to TechCrunch.

The numbers land in the middle of an agent boom

The figures arrive amid heavy promotion of a new generation of AI agents that transact on users' behalf rather than just answering questions. TechCrunch points to a wave of personal-agent startups, including Instinct, which raised its most recent funding round at a $10 billion valuation, and Meta's Muse assistant, which topped Apple's US App Store chart last month with more than 10 million users, per Meta's own figures, many of them reportedly shopping and transacting through it. OpenAI, Anthropic, Google and SpaceXAI (Grok) all offer agents as well.

Uber is still betting on agents

Despite the weak early numbers, Khosrowshahi was upbeat about the category. He said there are currently around eight major AI assistants on the market and expects that to grow beyond ten, adding that "every single AI system company is talking to us." He framed agents as another on-ramp to Uber's platform and predicted they would be a net positive for demand across its ecosystem.

Why it matters

Public, first-party transaction counts are rare in the debate over agentic AI, which is otherwise driven by demos, funding rounds and app-store rankings. Uber's four rides, echoed by DoorDash's few hundred lifetime agent orders, suggest that for fast, routine purchases, agents are not yet displacing apps, possibly because the apps are already quick enough that delegation saves no meaningful time. Yet both companies remain publicly committed to agent integrations, and if every major AI developer is indeed courting Uber, the plumbing for an agentic commerce layer is being built regardless. The gap between that infrastructure investment and current user behavior is the metric to watch over the next several quarters.