The pitch arrived first: let an AI agent book the ride, order dinner, shop for groceries and clear the day’s small chores. Meta’s Muse and OpenAI’s Dots launched in September amid precisely those expectations.¹

By late October, DoorDash offered the first reality check. CEO Tony Xu told investors that outside agents had generated only “several hundred orders in total” since going live — negligible against the billions of orders DoorDash handles each year.²

Uber’s Nov. 4 earnings call made the gap still starker. During Halloween week, the company completed 320 million rides and deliveries, one of its biggest weeks, but agents accounted for roughly 500 Uber Eats orders and just four ride-hailing trips. CFO Balaji Krishnamurthy called the response “very, very early” and said traction remained limited.²

The explanation may be brutally simple: the chores agents promise to automate are already quick. Uber CEO Dara Khosrowshahi said a typical Uber order takes 50 seconds and an Uber Eats order 90 seconds. Tech researcher Matt Maher argued that the industry is piling elaborate technology onto tasks that never demanded it: “Big Tech companies bludgeon us with solutions that are desperately looking for problems.”¹

Trust is another drag. Consumers remain wary of letting an autonomous tool choose purchases or handle card details, particularly without clarity about how their data is used.¹

Yet the weak third-party numbers have not cooled the platforms’ own AI ambitions. Uber says its Cart Assistance produces baskets roughly twice the average size; DoorDash is testing text-based ordering and its Ask DoorDash tool; Airbnb plans an in-house trip-planning agent. The companies are not closing the door on outside agents, either: Khosrowshahi said Uber is negotiating with AI providers and sees them as another route to its platform.¹

For now, though, the supposed AI shopping revolution is meeting a stubborn human response: ordering for yourself is still easier — and feels safer.