DoorDash CEO Says AI Agents Have Driven Only Hundreds of Orders

DoorDash co-founder and CEO Tony Xu told analysts that outside AI agents have generated only a few hundred orders on the platform since launching, a figure he called a "very low use case" for a company that processed roughly 1 billion orders in the third quarter of 2026 alone, according to Ars Technica.

Xu made the comment during DoorDash's third-quarter earnings call on Thursday, in response to an analyst question about agentic commerce. He said AI agents are "generating several hundred orders across the time that they've been up and running," per the call as reported by Ars Technica.

How many orders are AI agents actually driving?

The number Xu cited — several hundred — covers the entire period since outside AI agents gained access to DoorDash's ordering system. That is a rounding error against roughly 1 billion total orders placed on the platform in the third quarter of 2026, Ars Technica reported.

Xu's figure applies specifically to third-party agents, not DoorDash's own in-house AI tools. The company has not released a comparable usage count for those internal features.

Why does DoorDash's CEO call this a low-use case?

Xu drew a distinction between the kind of purchases he thinks AI agents are suited for and the kind DoorDash sells. He said he sees potential for agents in "long consideration purchases" — bigger decisions where an agent could work "proactively in the background" on a person's behalf.

Food delivery, he argued, does not fit that pattern.

"When it comes to a task that's as simple as ordering food, we tend to see that come directly to DoorDash," Xu said, according to Ars Technica's account of the earnings call.

What is DoorDash's MCP connector and who can use it?

Part of the low number may come down to timing. DoorDash rolled out a Model Context Protocol connector less than a month before the earnings call, Ars Technica reported. MCP is an open standard that lets third-party AI tools place orders through DoorDash's platform in a more direct way than scraping a web interface.

Access to that connector remains limited. Ars Technica reported it is available only to a small number of "early beta partners" and is mainly being used to streamline corporate lunch orders, not individual consumer purchases. Agents without that linkup have to navigate DoorDash's web interface, a more cumbersome path that likely caps adoption for now.

How does Ask DoorDash differ from third-party AI agents?

Xu's several-hundred-order figure does not include DoorDash's own AI features, chief among them a tool called Ask DoorDash that the company launched this summer. Ask DoorDash can build a shopping cart from a photo of a recipe or from a vague prompt such as "dinner for a family of four," according to Ars Technica.

DoorDash has not published a specific order count for Ask DoorDash, but the company told Ars Technica last month that baskets built through the tool come together five times faster than those built without it. Ask DoorDash orders also run nearly 50 percent higher in basket value and include about 60 percent more unique items than the same customers' traditional orders, the company said. Close to half of Ask DoorDash orders land at local restaurants the customer had never tried before, according to DoorDash's figures cited by Ars Technica.

That split matters for how the earnings-call numbers should be read. Xu's "low use case" comment describes outside agents built by other companies, not DoorDash's internal push to put AI closer to the ordering process itself.

What should readers expect next for AI agent shopping?

Ars Technica frames the current numbers as an early-stage snapshot rather than a verdict on agentic commerce broadly. The piece draws a comparison to Pizza Hut's first online ordering tests in 1994, when it would have been hard to picture a future in which more than half of Pizza Hut's orders arrive online.

Whether AI agents follow a similar trajectory for food delivery remains unproven. For now, the gap between hundreds of agent-driven orders and roughly a billion total quarterly orders is wide enough that DoorDash's own CEO is not predicting an imminent shift.

What to watch

• Whether DoorDash expands MCP connector access beyond early beta partners and corporate lunch accounts.
• Whether DoorDash publishes a specific order count for its internal Ask DoorDash tool.
• Whether other food-delivery or e-commerce platforms report similar low adoption numbers for outside AI agents in coming earnings calls.
• Whether basket-value and discovery metrics for AI-assisted orders continue to outpace traditional orders as volume grows.