All posts
2 min readby Romiel Inolino

DoorDash handled just under 1 billion orders last quarter. Outside AI agents placed several hundred in total

AI agentsecommerceMCPDoorDashCRM

The agent commerce wave is real in press releases. In DoorDash's order book, it is several hundred orders.

What happened

On DoorDash's third-quarter earnings call, CEO Tony Xu said third-party AI agents have generated "several hundred orders" in total since they started, Ars Technica and TechCrunch report. DoorDash handled just shy of 1 billion orders in the third quarter alone, per Ars. Xu called it a very low use case and said: "you never want to fight gravity. But at the same time, you don't want to be delusional."

Context matters. DoorDash's MCP connector for outside AI tools launched less than a month ago, is limited to a small number of early beta partners and is mainly used for corporate lunch orders. Xu said agents may suit long consideration purchases, but simple tasks like ordering food tend to come straight to DoorDash.

DoorDash's own AI tools are a different story by the company's account. Ask DoorDash, launched in June, builds baskets five times faster, and those orders average nearly 50% higher basket value than the same customers' usual orders, the company said last month. It also runs a text-to-order agent and a CLI agent. TechCrunch notes Amazon has been blocking personal shopping agents while Walmart welcomes them.

My take

For the small and mid-sized businesses I build for, this is a useful priority check. The volume is not coming from outside agents yet. The measurable lift DoorDash reports comes from AI inside its own channel, helping a customer who is already there.

So my order of work for clients:

  1. AI where your customers already talk to you: website chat that books, SMS that takes the order, follow-ups from the CRM.
  2. Clean, structured product and service data, which helps both your own assistant and any outside agent later.
  3. Only then, an MCP or agent endpoint for outside tools, with an owner and a metric.

Track agent-originated orders as their own source in the CRM now. When that number moves, you will see it before your competitors do.

I build in-channel assistants and CRM handoffs for teams. Examples at romielwillautomate.dev.

More posts