What Is an AI Agent, Really?
Before diving into the corporate world, let's clarify the term. An AI agent isn't just a chatbot. It's a system designed to understand a high-level goal, break it into steps, and then take action to achieve it with minimal human input. Think of it as
moving beyond just answering a question ('What's a good recipe?') to executing a task ('Order the ingredients for this recipe'). The hype suggests these agents will soon book our trips, order our meals, and manage our schedules. However, a significant gap exists between this vision and the current state of enterprise adoption, where most projects are still in early phases.
The Sobering Reality of Third-Party Agents
Recent earnings calls from these platform giants have painted a stark picture of consumer adoption for external AI agents—like those from Meta or OpenAI—that promise to interact with services on a user's behalf. During Halloween week, a peak time with 320 million transactions, Uber reported that AI agents booked a grand total of four rides and about 500 Uber Eats orders. Similarly, DoorDash's CEO noted that agents have generated only 'several hundred orders' in total, and Airbnb's CEO described bookings from such agents as 'very non-material'. The numbers are statistically negligible, suggesting the revolution where users bypass apps in favor of a universal AI assistant hasn't arrived. The simple reason may be that ordering an Uber already takes less than a minute, a process that isn't broken and doesn't need an AI middleman to fix.
DoorDash: AI for Efficiency and Discovery
While outside agents see little traction, DoorDash is aggressively integrating its own AI. The company uses AI agents across its entire operation, serving customers, delivery drivers ('Dashers'), and merchants. One major area is automating the once-tedious process of placing phone orders with restaurants not fully integrated into its app, where AI now handles over 35,000 orders daily. For customers, the 'Ask DoorDash' chatbot helps build grocery and food orders based on conversational prompts, with early tests showing users complete carts five times faster. The company is also using sophisticated AI simulation frameworks to test and improve its customer service chatbots, successfully reducing errors and 'hallucinations' before they reach users. This is a story of internal, targeted optimization, not a futuristic, all-encompassing agent.
Uber: Internal Tools and Cautious Optimism
Like DoorDash, Uber's most significant AI agent usage is internal. The company has heavily invested in AI agents for its software engineering teams, where they handle tasks like code reviews, debugging, and fixing system failures. This has led to massive efficiency gains, with over 70% of code pull requests being attributed to these internal agents. On the customer-facing side, the approach is more measured. Uber Eats has a feature called 'Cart Assistance', which allows a user to upload a photo of a recipe or a handwritten grocery list, and an AI fills their cart. While the volume is low, these AI-assisted baskets are roughly twice the size of an average order, showing a clear business case. Uber remains open to third-party agents but acknowledges that for now, their impact is minimal.
Airbnb: Rejecting Hype for a Custom Experience
Airbnb is taking the most distinct path. CEO Brian Chesky has openly stated that generic, all-purpose AI agents are not suitable for travel planning, an activity he says is aspirational and part of the fun for 80% of travelers. Instead of a simple chatbot, Airbnb is building its own proprietary AI travel agent, set for a 2027 launch. This agent will use 'Generative UI' to build a custom, visual interface for each user with maps, photos, and comparisons, rather than just text. Crucially, it will also handle post-booking tasks like refunds, rebooking, and safety support—complex actions that require deep integration and that third-party agents cannot perform. Airbnb is betting that specialized, in-house AI that enhances the user journey will win over generic agents that try to replace it.
















