The Promise of an AI Co-Pilot
For the past year, major tech platforms have been selling a compelling vision of the future. Instead of tapping through menus and filters, you’ll simply talk to an AI 'agent'. This digital assistant will understand your needs—like 'find a spicy vegetarian
meal for two' or 'book a pet-friendly apartment near the beach for next weekend'—and execute the multi-step task for you. Companies like DoorDash, Uber, and Airbnb have all invested heavily in this concept. DoorDash launched 'Ask DoorDash', a conversational tool to build grocery carts from a photo of a recipe. Uber is exploring AI agents for everything from trip planning to internal coding. Airbnb, meanwhile, is set to roll out its own AI travel agent in 2027, aiming to handle everything from initial booking to post-trip support. The goal is to create a seamless, intuitive experience that feels less like using an app and more like having a personal assistant.
A Reality Check from the Front Lines
Amid the hype, a recent earnings call from DoorDash provided a dose of reality. CEO Tony Xu revealed that third-party AI agents—external services like Muse and Instinct that can place orders on a user's behalf—have been responsible for only 'several hundred orders in total'. To put that into perspective, DoorDash handles millions of deliveries every single day. Uber reported a similarly quiet start, with third-party agents booking just four rides and around 500 Uber Eats orders in a recent quarter. Xu’s commentary was measured, acknowledging the potential of the technology while cautioning against getting carried away. 'My take on all things technology is you never want to fight gravity. But at the same time, you don't want to be delusional,' he stated. This small figure isn't a sign of failure, but rather a crucial indicator of the real-world friction involved in changing user behaviour and perfecting the technology.
The Hurdles: Why So Few Orders?
The gap between the AI vision and current adoption highlights several significant challenges. First is the technical complexity. Building an AI agent that can reliably interpret ambiguous human language and navigate an app's interface without error is incredibly difficult. A mistake in a search query is frustrating; a mistake that orders the wrong, allergy-inducing meal is a serious problem. Second, there is the issue of consumer trust and habit. Users are accustomed to the control and predictability of tapping through an app. Handing over purchasing power to an AI, especially with personal payment information, requires a leap of faith that many are not yet willing to take. Finally, there's the question of utility. For simple, routine tasks, many users find it faster to use the existing app interface than to articulate their request to a chatbot. As Xu noted, customers are still coming directly to the DoorDash app for their everyday orders because it works.
The Long Game: Crawl, Walk, Run
Despite the low initial numbers from third-party agents, these companies are not backing down. They are playing a long game, viewing this initial phase as a critical period of testing and data collection. The 'several hundred' orders, while trivial in volume, provide invaluable data on how these agents operate in the wild. Both DoorDash and Uber are simultaneously developing their own in-house AI tools. DoorDash has seen success with its 'Ask DoorDash' feature, noting that it builds grocery carts five times faster and results in basket values nearly 50% higher than conventional orders. Uber Eats has a similar feature that fills a shopping cart from a photo of a grocery list. Similarly, Airbnb has used AI to power its customer service bots, successfully resolving nearly 45% of issues without human intervention. This strategy allows them to control the user experience, ensure safety, and build trust gradually. The plan is to start small, refine the technology, and slowly get users comfortable with the idea of an AI-powered future.
















