In a nutshell: The promise that personal AI agents would revolutionize the way we use ride-hailing apps, order food, and book accommodation is looking a bit shaky right now. Leaders of these industries say almost nobody is using the likes of Meta's Muse or OpenAI's Dots to interact with their services. Uber, for example, said only four rides were ordered through an AI agent during the week of Halloween, its busiest week in October.

Uber, DoorDash, and Airbnb have all said that AI agents are responsible for only hundreds of transactions on their platforms, a tiny fraction compared to the millions completed by real people daily.

Uber chief financial office Balaji Krishnamurthy said the company took around 320 million trips and orders across Uber and Uber Eats during the week of Halloween. In addition to just four rides being placed by AI agents, only 500 Uber Eats orders were placed by these AIs.

"It is very, very early and very limited traction that we are seeing so far," said CEO Dara Khosrowshahi. "Listen, we hope that there will be more traction."

DoorDash CEO Tony Xu said it was a similar story at the food delivery company. He told analysts that AI agents have generated just several hundred orders in total since they became available. Muse launched on September 8, while Dots launched on September 29. DoorDash receives billions of orders globally each year.

More AI companies are offering personal agents, including Anthropic, Google, Instinct, and SpaceXAI. Muse was even top of the US App Store last month. But some say they are solutions in search of a problem.

In the case of Uber, Khosrowshahi pointed out that it takes most people around 50 seconds to book a ride and 90 seconds to reorder their favorite food. It's hard to imagine many people so pushed for time that they decide using an AI agent would be a more optimal method.

Airbnb CEO Brian Chesky, meanwhile, said that of its nearly half a million daily bookings, only a negligible number are made through AI agents such as Muse and Instinct. Chesky pointed out most people like to research and plan their trips themselves, checking out which areas to stay in, what sights to see, and what the homes look like, rather than letting an AI make the choices.

Not many users want to put their faith in an AI, either. There have been plenty of stories about AI agents performing tasks they weren't supposed to – or going about it in a less-than-ethical way. One recent example was that of an Australian user who asked an agent to book a gym class. Wanting to get the user in quickly, it found a weakness in the booking platform's API, removed the person at the top of the class waitlist, and replaced them with the user.