The tech world is buzzing about AI agents that can act on your behalf. Giants like Uber, DoorDash, and Airbnb are all in. But with booking already taking mere seconds, what problem is this technology actually trying to solve?
What is an AI Agent, Really?
First, let's clear things
up: an AI agent is not just a chatbot. While a chatbot can answer questions, an AI agent is designed to understand a goal and then autonomously perform the multi-step tasks required to achieve it. Think of it less like asking a help desk for information and more like giving a personal assistant a complex instruction, like 'Book me a quiet hotel near the conference centre for next Tuesday'. The agent then works through the process of searching, comparing, and even booking without you needing to manage each step. These systems are built to act, not just advise, which is why they represent a significant shift from the AI tools we've used before.
The Sobering Reality of Early Adoption
Despite the hype, the initial uptake of third-party AI agents for services like Uber, DoorDash, and Airbnb has been surprisingly low. Uber reported that during a busy Halloween week with 320 million trips and deliveries, AI agents were used to book just four rides and about 500 Uber Eats orders. Similarly, DoorDash's CEO noted that agents have generated only a few hundred orders in total since their launch. The reason seems simple: these apps are already incredibly efficient. Uber's CEO pointed out that booking a ride takes a median of 50 seconds, while reordering a favorite meal on Uber Eats takes 90 seconds. For simple, habitual tasks, asking an AI agent can feel like an unnecessary extra step.
Uber and DoorDash: AI for Complexity, Not Speed
So, if not for speed, what are these companies using AI for? Both Uber and DoorDash are developing their own in-house AI tools to handle more complex or nuanced requests. For instance, Uber Eats has a feature called Cart Assistance, where you can upload a photo of a recipe, and AI fills your shopping cart with the ingredients. Baskets made this way are reportedly twice the average size. DoorDash is experimenting with a text-to-order agent that can understand vague requests like 'order my usual' or handle complicated group orders with mixed dietary needs. These features aren't about shaving seconds off a simple order; they're about tackling tasks that would otherwise be tedious or difficult for a user to do manually.
Airbnb's Vision: The AI-Powered Trip Planner
Airbnb's ambition for AI is perhaps the most comprehensive. The company is building its own AI agent, with a planned rollout in the first half of 2027, designed to be a complete trip planner. The goal is to move beyond simple bookings and help users plan entire itineraries, from finding the right stay to suggesting local experiences. CEO Brian Chesky has noted that travel planning is a highly visual and personal process that people often enjoy, which is why a generic chatbot isn't enough. Airbnb's agent will use a dynamic interface with maps, photos, and comparisons, assembled in real-time based on the user's conversation. The focus is on discovery and planning, tasks where AI can add significant value beyond what a simple booking form can offer.
The Real Goal: Personalisation, Efficiency, and Control
The integration of AI agents is ultimately about more than user convenience. For these platforms, the real wins are in deeper personalisation, operational efficiency, and maintaining control of the customer relationship. AI can analyse vast amounts of user data to provide hyper-personalised recommendations, increasing the chances of a user making a purchase. On the backend, AI is a powerful tool for reducing costs. For example, about 40% of Airbnb's customer support issues are now resolved without a human agent, significantly cutting down on support costs per booking. By building their own proprietary agents, these companies also ensure they own the entire user journey, from initial query to final payment, preventing third-party agents from turning their platforms into mere fulfillment layers.
















