Companies

Four Rides: Uber Says AI Agents Are Nowhere Near Taking Over

Uber disclosed that third-party AI agents booked just four rides out of 320 million transactions last quarter, plus roughly 500 Uber Eats orders around Halloween.

Uber only fulfilled 4 rides from AI agents last quarter
Uber only fulfilled 4 rides from AI agents last quarterJESHOOTS-com / pixabay
By Elena Vasquez6 min read

Why it matters

  • Only 4 rides were booked by third-party AI agents on Uber in Q3 2026, out of 320 million total rides and deliveries.
  • Uber recorded 'around 500' AI-agent Uber Eats orders during the week of Halloween.
  • This is the first time Uber has disclosed third-party AI agent transaction counts.
  • CEO Dara Khosrowshahi said Uber will 'work with these third parties' but wants to be 'not dependent on any one player.'
  • Uber rolled out an AI Cart Assistant for Uber Eats earlier this year that builds grocery lists from recipe images or text prompts.

Uber fulfilled just four rides booked by third-party AI agents last quarter. Four. Out of roughly 320 million combined rides and Uber Eats deliveries in the third quarter of 2026, according to figures Uber executives disclosed on an earnings call this morning. The food side did marginally better: Uber recorded "around 500" Uber Eats orders placed by AI agents during the week of Halloween alone.

The numbers land amid real anxiety. This is the first time Uber has ever disclosed how many third-party AI agents have arranged rides or deliveries on its platform, and the disclosure comes at a moment when the industry is actively debating whether such agents could disintermediate the big consumer apps entirely.

Why four rides out of 320 million matters

The fear has a name: disintermediation. With the rise of AI agents like Meta's Muse or OpenAI's Dots, some experts now argue that platforms like Uber and DoorDash face a structural threat. In that scenario, a customer asks an AI agent to book a ride or order dinner, and the agent — not the customer — interacts with the marketplace. The app becomes plumbing. The brand relationship, and the pricing power that comes with it, migrates to whoever owns the agent.

That is a genuinely existential question for companies whose entire business is being the interface between consumers and gig workers. It's why analysts and investors keep pressing platform executives on it. And it's why Uber's answer — four rides, about 500 food orders — is the most concrete data point the industry has produced so far on whether agent-driven disintermediation is actually happening at scale.

The answer, at least for now, is that it isn't.

Executives framed the figures not as a ceiling but as a starting line. During the call, Uber's leadership expressed openness to working with third-party AI companies to make the platform easier for agents to use. At the same time, they voiced confidence that customers will keep preferring the Uber app itself.

What Dara Khosrowshahi actually said

Uber CEO Dara Khosrowshahi did not treat the agent question as a threat to bat away. He treated it as an early-stage negotiation.

"We hope that there will be more traction," Khosrowshahi said during the call. "We do think that we want to make services that are available to consumers in a delightful, easy way. We think our direct services are going to be the most delightful and easy, but we will work with these third parties."

That is a careful two-track position. Uber wants to be wherever the customer is, including inside someone else's agent. But it is explicitly betting that its own app will remain the better experience, and it intends to keep it that way.

He continued: "It is just very, very early, and we want to make sure we are not dependent on any one player, and I think we are very well positioned in this environment going forward."

The phrase "not dependent on any one player" is the tell. Whatever deals Uber eventually strikes with AI companies, it plans to avoid handing any single agent exclusive access to its supply of drivers and couriers. In a market where a handful of firms — Meta and OpenAI among them — are racing to own the consumer's AI assistant, that hedging instinct is unsurprising.

How big is the disintermediation risk, really?

The disclosed numbers put hard data behind what early adopters already know: this first wave of AI agents still has a long way to go before people see them as a viable way to order food or hail a ride. The friction points remain numerous. Booking a car through a conversational agent is, for most users today, slower and less reliable than tapping a button in an app that already knows your location, payment method and preferences.

The incentive structure cuts the same way. Platform operators like Uber and DoorDash have little reason to let outside agents barge in and take over their customer relationships. Every transaction that starts in someone else's agent is a transaction where the platform risks becoming an interchangeable commodity backend — exactly the fate Uber's disclosure suggests it intends to avoid.

None of this means the AI companies will stand still. As industry observers have noted, these companies know that commerce and transactions are among their agents' primary use cases — arguably their best path to a more profitable future, since actually completing purchases is where agents generate defensible value rather than just answering questions. They will keep pushing.

But getting Uber to agree to meaningful disintermediation will be a tough ask. The company just told the market, with numbers attached, that agents represent a rounding error in its business. That gives it years of leverage in any negotiation over how, and how generously, third-party agents get access.

Uber is building its own AI instead

Uber is not waiting passively to find out how the agent era unfolds. The company is developing its own AI tools, and their purpose is explicit: keep customers inside Uber's own platform rather than someone else's interface.

The most concrete example came earlier this year, when Uber rolled out a new Cart Assistant for Uber Eats. The feature puts an AI bot to work on grocery ordering: it builds a grocery list from an image of a recipe or from a text prompt. Snap a photo of a dish you want to cook, and the assistant assembles the shopping list for you — inside the Uber Eats app, not inside a third-party agent.

That is the strategic logic in miniature. If AI agents are going to mediate how people buy things, Uber would rather that agent be its own. The Cart Assistant competes for the same interaction — natural-language, intent-driven shopping — that Meta's Muse or OpenAI's Dots would need to win in order to disintermediate the platform.

What happens next

The four-ride figure will age quickly either way. If agent bookings grow from hundreds per week into millions, Uber's "very, very early" framing gets tested, and the balance of power between platforms and AI companies shifts. If growth stays flat, Uber's bet on its own app — and its own AI — looks prescient.

For now, the company has set the baseline. Every future disclosure of agent-driven transaction volume will be measured against four rides and 500 Halloween-week food orders. And Uber has made clear it plans to be a participant in the agent economy on its own terms: open to third parties, confident in its app, and determined not to let any single AI company own the customer.

Source: The Verge AI

Share this article:

More from Elena Vasquez

Elena Vasquez

Show full bio

Market editor covering technology and product news in artificial intelligence and machine learning at AI In Context.

379 articles

Related articles

  1. Meta Claims Millions Shop With Muse. Executives See 'Hundreds' of Orders
  2. Meta's Muse Agent Hits 5 Million Downloads, Lifts Stock 17%
  3. OpenAI's $100-a-Month Dots Agent Misheard Me, Said It Loved Me
  4. Multi-Agent AI Teams Cost Up to 5.1x More for Marginal Gains
  5. DoorDash has seen only 'several hundred' orders from AI agents

« Previous article