A Drop in the Ocean
During a late October 2026 earnings call, DoorDash CEO Tony Xu revealed that third-party AI agents have been responsible for just 'several hundred' orders in total since the feature became available. To put that number in perspective, DoorDash processes
millions of orders every single day, and handled nearly a billion in the third quarter of 2026 alone. This makes the volume from external AI agents—software like Meta's Muse or other assistants that can autonomously place an order on a user's behalf—a statistical rounding error. The disclosure offers a rare, concrete data point showing the massive gulf between what AI can do in a demonstration and what it is actually doing in the real world of consumer commerce.
The Human Habit Hurdle
So, why the slow start? One major factor is that these AI agents are often trying to solve a problem that doesn't really exist for most people. Ordering food on an app like DoorDash is already incredibly simple and fast; Uber's CEO noted it takes an average of just 90 seconds to order from Uber Eats. For many routine tasks, it's simply easier for a person to open the app and tap a few buttons themselves rather than delegating the task to an AI. Beyond convenience, there are issues of trust and reliability. Some users are cautious about letting an autonomous agent 'go rogue' with their credit card details, especially given early reports of ordering errors with similar technologies. The adoption of any new interface, from mobile apps to voice commands, has always required time for users to build new habits, and AI agents haven't yet found a compelling reason for most people to make the switch.
DoorDash's Cautious Optimism
Despite the low numbers, DoorDash isn't writing off AI. CEO Tony Xu struck a balanced tone, stating, 'you never want to fight gravity. But at the same time, you don't want to be delusional.' His comments distinguish between the limited use of third-party agents and the potential of AI overall. In fact, DoorDash has been actively developing its own AI tools, such as the 'Ask DoorDash' chatbot and a text-to-order feature. The company has suggested its own internal AI tools lead to higher-value orders from customers discovering new local restaurants. By keeping the door open to external agents while building its own, DoorDash is positioning itself to learn about user behaviour and potentially find future revenue streams, like advertising within those third-party agents, should they ever gain traction.
A Reality Check for the AI Revolution
DoorDash isn't alone. Executives at Uber and Airbnb have reported similarly negligible numbers of transactions coming from external AI agents. Uber's CFO noted that during its busy Halloween week, AI agents booked just four rides and about 500 Uber Eats orders out of 320 million total trips. The consensus is that while agents might be good for simple, repetitive tasks, they have yet to prove their worth for more complex or personal decisions, like planning a trip. This situation highlights a broader strategic dilemma for platforms. Some, like Walmart, are rolling out the red carpet for agents to potentially steal business from competitors, while others, like Amazon, are reportedly blocking them. For now, the data from DoorDash suggests that the agent-led commerce revolution is more of a slow burn than a sudden explosion.
















