Franklin AI News Brief

DoorDash says third-party AI agents have placed only several hundred orders

Key Takeaways

  • The company’s reported order count puts current agent commerce in perspective without predicting its eventual scale.
  • A small observed channel DoorDash says it has received only several hundred orders in total from third-party AI agents, according to TechCrunch’s coverage of its third-quarter earnings call.
  • CEO Tony Xu describes the area as dynamic while saying customers still predominantly order through the company’s app.
  • That distinction is useful amid predictions about agents taking over online commerce.
  • The reported total describes activity DoorDash has observed from particular outside agents.

A small observed channel

DoorDash says it has received only several hundred orders in total from third-party AI agents, according to TechCrunch’s coverage of its third-quarter earnings call. CEO Tony Xu describes the area as dynamic while saying customers still predominantly order through the company’s app.

That distinction is useful amid predictions about agents taking over online commerce. The reported total describes activity DoorDash has observed from particular outside agents. It does not establish the size of every agent-mediated shopping market or show how consumers will behave once different integrations become available.

Keep the comparison periods clear

TechCrunch contrasts the agent figure with DoorDash’s 2025 report of an average three million deliveries per day. One number is a reported total for third-party agent orders; the other is an earlier daily company-wide average. They provide context, but they are not matched periods from which to calculate a precise current market share.

DoorDash is also developing its own interfaces. The report lists a text-to-order agent, a command-line agent, Ask DoorDash and an MCP connection for corporate ordering through other chatbots. Those first-party capabilities are distinct from the outside-agent order count.

Interest does not equal established distribution

Xu says DoorDash wants to learn from agent use cases and develop relationships with third-party providers. The report also describes a potential future advertising opportunity within agents. That is an option being discussed, rather than evidence of an already material advertising business from this channel.

For teams building agent commerce, the actionable distinction is between an integration being technically possible and customers choosing to complete transactions through it. Look for completed orders, repeat usage and the specific permission model, alongside any demonstration of an assistant filling out a shopping request.

Our read

Franklin AI Take

DoorDash’s count concerns third-party agents, while the company is also developing its own ordering interfaces. Keep those channels separate when assessing adoption; a working integration alone does not show sustained transaction demand. Completed repeat orders are the more useful test of adoption than the availability of an integration alone.