The Promise of an AI Co-Pilot
Imagine telling your phone to just 'order your usual Friday night meal' or 'find a family-friendly holiday spot for next summer'. This is the world that companies like Uber, DoorDash, and Airbnb are building towards with their new AI-powered assistants.
DoorDash, for instance, has been piloting an AI assistant that helps you discover meals based on a 'vibe', build a grocery cart from a recipe, or quickly reorder favorites. Uber has introduced AI features to help customers, including a 'Cart Assistance' for Uber Eats that can fill your basket from a handwritten grocery list. Meanwhile, Airbnb has announced plans for a sophisticated AI agent, set to roll out in 2027, that goes beyond a simple chatbot to help users plan and book entire trips with a dynamic, visual interface.
A Reality Check on Adoption
Despite the significant investment and hype, the initial uptake of these third-party AI agents has been strikingly low. During Halloween week, a peak time for the company, Uber reported that AI agents booked just four rides and around 500 Uber Eats deliveries out of a combined 320 million transactions. DoorDash CEO Tony Xu noted a similar trend, stating that external AI agents had generated only 'several hundred orders' in total since their launch. While these companies are still hopeful, the numbers paint a clear picture: consumers are not yet flocking to delegate their shopping and booking tasks to these new AI middlemen.
A Solution in Search of a Problem?
So, why the disconnect? For many users, these AI agents seem to be overcomplicating tasks that were already simple. Uber's CEO, Dara Khosrowshahi, pointed out that it already takes a median of just 50 seconds to book a ride and 90 seconds to place a food order. For a task that quick, introducing an AI agent can feel like adding an unnecessary step. Furthermore, consumer trust remains a significant barrier. Surveys show that a large percentage of shoppers are wary of AI, with concerns about data privacy, a lack of need for the tools, and a preference for human assistance. One 2025 survey found that 54% of consumers simply don't see the need for AI shopping assistants in the first place.
The Human Element of Shopping
Part of the slow adoption may also stem from the nature of the tasks themselves. Booking a trip, as Airbnb's CEO Brian Chesky has noted, is an aspirational and often enjoyable process of discovery that many people don't want to fully automate. Users enjoy browsing photos, reading reviews, and comparing options—activities that a purely transactional AI might strip away. This suggests that the most successful AI integrations will be those that assist, rather than replace, the user's decision-making process. While a machine can be told to 'book a flight', it can't replicate the joy of finding the perfect holiday rental. Trust is also a major factor, with surveys showing that only a small fraction of consumers would let an AI make a final purchase decision for them.
Early Days or a Fundamental Flaw?
Despite the low numbers from third-party agents, these companies aren't giving up on AI. Instead, they are focusing on building AI features directly into their own apps, where they have more control and can solve more specific problems. Uber's 'Cart Assistance' feature has seen success, with AI-assisted baskets being roughly twice the size of average orders. DoorDash is also experimenting with its own in-app conversational search. This shift suggests the future of AI in these services may be less about all-powerful external agents and more about smart, integrated features that enhance the existing user experience. It's not about replacing the app, but making the app better.
















