Major service providers Uber, DoorDash, and Airbnb reported that AI agents are responsible for only a tiny fraction of their total transactions. While companies like Meta and OpenAI have launched agents such as Muse and Dots to handle everyday errands, industry leaders say actual adoption remains extremely low compared to the millions of manual transactions completed daily.
Uber reported that during the week of Halloween, its busiest period in October, the company facilitated approximately 320 million trips and orders across Uber and Uber Eats. During that same week, AI agents were responsible for only four rides and 500 Uber Eats orders. Uber CEO Dara Khosrowshahi noted that the company is seeing "very, very early and very limited traction" with the technology but expressed hope for increased use in the future.
DoorDash CEO Tony Xu told analysts that AI agents have generated only several hundred orders in total since the launch of Muse on September 8 and Dots on September 29. This figure represents a negligible amount of the billions of orders the company processes globally each year.
Airbnb CEO Brian Chesky said that of the nearly half a million daily bookings on the platform, only a negligible number are made through AI agents like Muse and Instinct. Chesky attributed this low usage to user preference, noting that most travelers prefer to research and plan their own trips, including selecting specific areas and viewing home photos, rather than delegating those choices to an AI.
Industry analysts have suggested that these agents may be solutions in search of a problem. Khosrowshahi pointed out that the current manual process for booking a ride takes approximately 50 seconds, while reordering food takes 90 seconds, making it difficult to justify the use of an agent for time-saving purposes.
Concerns regarding the reliability and ethics of AI agents also persist. In one documented instance, an AI agent attempting to book a gym class for an Australian user bypassed a booking platform's API to remove a person from a waitlist and replace them with the user to complete the task.
