The AI Agent Promise vs. Reality
For months, the tech world has buzzed with the promise of 'agentic AI'—proactive digital assistants that do more than just answer questions. Unlike a simple chatbot, an AI agent can take action on your behalf, like booking a reservation or completing
a purchase. Tech giants have rolled out high-profile agents, predicting a future where users delegate daily tasks to software. However, recent earnings calls from Uber, DoorDash, and Airbnb paint a very different picture of early consumer adoption, revealing that the revolution has been slow to start.
Uber: 'Very Limited Traction'
Uber's executives provided some of the most concrete, and startling, figures. During Halloween week, one of the company's busiest periods, third-party AI agents booked a grand total of four Uber rides and around 500 Uber Eats deliveries. This is out of approximately 320 million combined transactions on the platform during that same week. CEO Dara Khosrowshahi described the numbers as reflecting 'very, very early and very limited traction'. He suggested a simple reason: efficiency. Ordering an Uber already takes a median of just 50 seconds, leaving little friction for an AI to solve. Still, Uber isn't dismissing the technology. The company is developing its own tools, like a 'Cart Assistance' feature for groceries that has resulted in shopping baskets twice the average size. Khosrowshahi also confirmed Uber is in discussions with all major AI system companies, viewing them as a potential new channel for customer acquisition rather than a threat.
DoorDash: Hundreds, Not Millions
DoorDash CEO Tony Xu reported a similar trend, stating that external AI agents have generated 'maybe several hundred orders in total' since their launch. Considering DoorDash handles millions of deliveries daily, this volume is negligible. Like Uber's chief, Xu is pragmatic, acknowledging that while you 'never want to fight gravity' with technology, you also 'don't want to be delusional'. The company is hedging its bets by developing its own AI capabilities. It has launched an 'Ask DoorDash' chatbot and even a text-to-order service, allowing users to place orders via natural language and photo uploads within the DoorDash ecosystem. This dual strategy—engaging with third-party agents while building a proprietary experience—shows the company is preparing for an AI-driven future, even if it hasn't arrived yet.
Airbnb: The Joy of the Search
Airbnb's situation is unique. CEO Brian Chesky confirmed that a 'very non-material number' of bookings come from external AI agents. He attributed this not to a failure of technology, but to user behaviour. Planning a trip, he argued, is an aspirational and enjoyable activity that people don't necessarily want to automate. He pointed out that users often spend work hours browsing dozens of listings and scrolling through photos—an experience that is personal and visual. While skeptical of general-purpose agents for travel planning, Chesky is heavily investing in specialized AI. Airbnb's own AI assistant for customer service is already resolving over 40% of issues without human intervention, leading to a significant drop in support costs. The company is now building its own travel-specific agent, slated for a 2027 rollout, designed to help users plan their entire trip within the Airbnb app.
The Road Ahead for AI Agents
The message from these three consumer tech leaders is clear: the hype around third-party AI agents taking over commerce has far outpaced reality. Consumers are not yet ready or willing to hand over control for tasks that are already simple or enjoyable to do themselves. This doesn't mean the agentic era is a bust. Instead, it suggests a more nuanced evolution. These companies are actively negotiating the 'terms of trade' with AI developers while simultaneously investing heavily in their own, more integrated AI tools that enhance their existing platforms. The early data shows that AI is succeeding when it solves complex problems like customer service or enhances in-app discovery, rather than simply replacing a few taps on a screen.
















