Uber has put some hard numbers on one of Silicon Valley’s most breathless promises: AI agents are barely moving the needle on its platform.

During the Halloween week, its busiest stretch in October, third-party agents arranged just four rides and about 500 Uber Eats orders, against roughly 320 million combined rides and deliveries. The disclosure was Uber’s first public accounting of agent-driven transactions and offered a sharp reality check for predictions that personal assistants could pull customers away from apps such as Uber and DoorDash.¹

The timing matters. AI companies have marketed a new generation of assistants as tools that will not merely answer questions but complete purchases and bookings. Yet Uber’s figures suggest consumers remain reluctant to hand over even routine choices—calling a car or reordering dinner—to an intermediary bot.

On the company’s third-quarter earnings call, CEO Dara Khosrowshahi did not dismiss the model. “It is very, very early and very limited traction that we are seeing so far,” he said, adding: “Listen, we hope that there will be more traction.”²

Uber’s position is pragmatic rather than defensive. Khosrowshahi said agent access could create incremental demand and noted that major AI companies are already talking with Uber. But the company also has little reason to make itself invisible behind someone else’s assistant. Its argument is that the direct experience is already quick: an average ride can be booked in about 50 seconds, while reordering a favorite meal takes around 90 seconds.

That leaves the central tension unresolved. AI agents may eventually become a meaningful new front door to Uber, and the company says it will work with third parties. For now, however, the supposed disintermediation threat looks more like a pilot project than a consumer revolt. Uber is also building its own tools, including an Eats Cart Assistant that can turn a recipe image or prompt into a grocery list—an effort to keep AI convenience inside Uber’s walls.