An Almost Comical Statistic
During its third-quarter earnings call, Uber revealed a statistic that felt more like a punchline than a performance metric. In the week of Halloween, one of its busiest periods with approximately 320 million total trips and deliveries, external AI agents
were used to book a grand total of four rides. The numbers for its food delivery service were slightly more robust but still statistically tiny, with around 500 Uber Eats orders initiated by AI agents. In a tech landscape saturated with talk of AI disruption, these figures provide a stark, almost comical, dose of reality. For all the billions invested and the headlines generated by AI assistants, their actual impact on consumer behaviour for a service like Uber remains, for now, virtually zero.
What Are 'AI Agents' Anyway?
It's important to clarify what we mean by an 'AI agent' in this context. These are not the AI tools Uber is building into its own app, like a voice assistant to help you book a ride hands-free. Instead, the four bookings refer to transactions initiated by third-party personal AI assistants—think of sophisticated versions of Siri or Alexa, like those from OpenAI, Meta, and buzzy startups, designed to perform tasks for you across different apps and services. The dream is that you could simply tell your AI, 'Get me an Uber to the airport', and it would handle the entire booking process. However, as Uber's data shows, and as corroborated by similar reports from DoorDash and Airbnb, we are a very long way from that dream becoming a common reality.
Hype Meets a Practical Barrier
So why the massive disconnect? Uber CEO Dara Khosrowshahi offered a simple, practical explanation: the current Uber app is already extremely efficient. He noted that the average time it takes to book a ride is just 50 seconds, and re-ordering a favorite meal takes about 90 seconds. For a task this quick, delegating it to an AI agent doesn't necessarily save time or reduce friction. In fact, it might add complexity. This highlights a critical lesson for the entire tech industry: for AI to be widely adopted, it must solve a real problem more effectively than the existing solution. Right now, for booking a ride, the old-fashioned way of tapping a few buttons on a screen is winning by a landslide.
A Calculated Strategy, Not a Failure
While the low number is striking, Uber's leadership isn't panicking. Khosrowshahi, who previously ran Expedia, is wary of becoming dependent on outside 'gatekeepers' like Google, which dominated travel search. By controlling access to its platform, Uber can negotiate favorable terms with the many AI companies—Khosrowshahi counts at least eight major ones—that want to integrate with its service. The company sees AI agents as a potential new channel for demand, but one that it must control. This isn't a sign of failure but of a deliberate strategy to avoid repeating the power dynamics of the last tech wave, where platforms that owned the customer relationship held all the cards.
The Sobering Reality of AI Adoption
Uber’s report is a powerful reality check for the entire AI sector. It demonstrates that consumer habits change far more slowly than technology does. Issues like security, privacy, and simple trust remain significant barriers to users handing over their passwords and credit cards to an autonomous agent. While AI is making transformative strides in back-end operations and enterprise software, its role in everyday consumer life is still in its infancy. The four-ride week is a reminder that building the technology is only the first step; convincing millions of people that it's genuinely better, safer, and more convenient is the real challenge.
















