The Hype Meets a Quiet Reality
The promise of AI 'agents'—autonomous programs that can book, buy, and plan on your behalf—has been a dominant tech narrative. Yet, the first real-world data from the world's biggest platforms tells a different story. During Halloween week, Uber fulfilled
a staggering 320 million trips and deliveries, but external AI agents booked a grand total of four rides and about 500 Uber Eats orders. The numbers were so low that executives called the traction 'very, very limited'.This isn't an isolated case. DoorDash's CEO reported that AI agents had generated only 'several hundred orders' in total since becoming available. Meanwhile, Airbnb CEO Brian Chesky described bookings from these agents as 'very non-material' compared to the platform's nearly 500,000 daily bookings. The data suggests that for now, consumers are not outsourcing their ride-hailing and dinner plans to AI. The reason may be simple: convenience. Ordering an Uber already takes a median of 50 seconds, a task so quick that an AI agent doesn't offer a significant time-saving advantage.
Internal vs. External AI Agents
To understand this trend, it is crucial to distinguish between two types of AI. The low usage numbers primarily refer to external, general-purpose AI agents, such as Meta's Muse, which are designed to perform a wide variety of tasks for users across different services. These are the agents that have struggled to gain a foothold in complex consumer transactions.However, this is not the whole story. Internally, these same companies are investing heavily in their own specialized AI agents. These are not general-purpose assistants but highly focused tools designed to improve specific parts of their own business, from behind-the-scenes software development to customer-facing features within their own apps. While you may not be asking a third-party AI to book your ride, the platforms themselves are using AI in ways you might not see, but that still have a major impact on their operations and your experience.
The In-House AI Arms Race
While rejecting a reliance on outside agents, each of these giants is building its own AI powerhouse. At Uber, AI agents are already transforming how the company writes software. These internal agents are handling repetitive coding tasks, upgrades, and bug fixes, contributing a significant and growing percentage of code changes with no human authoring. The goal is to free up human engineers to focus on more creative, high-impact work.DoorDash is experimenting with a new way for customers to interact with its service, launching a beta for a text-to-order AI agent that bypasses the app entirely. The tool can remember a user's 'usual' order, make recommendations, and handle group orders, all within a simple text message thread.Airbnb is playing the long game with the most ambitious project of all: a proprietary AI travel agent set to roll out in 2027. The company believes planning a trip is a visual, social, and complex process that a simple chatbot cannot handle. Its agent will combine maps, photos, calendars, and reviews into a dynamic interface, moving far beyond text to help users discover and book their stays, and even manage reservations after they are made.
The Battle for the Customer Relationship
The strategic thinking behind this in-house push is clear: these platforms do not want to become a commoditized background service. Handing over the customer interaction to an external AI agent risks losing the direct relationship, brand loyalty, and valuable data that are central to their business models. Uber's CEO, having previously led Expedia, explicitly stated his desire to avoid depending on tech giants like Google for customer access again.By building their own AI tools, companies like Uber, DoorDash, and Airbnb ensure they remain the primary destination for their respective services. They control the user experience, gather the data, and capture the full value of the transaction. The strategy is not to resist AI, but to integrate it on their own terms, creating specialized tools that enhance their core product rather than allowing a third party to sit between them and their users. They are positioning themselves to set the rules for how AI interacts with their marketplaces.
















