Only hundreds of orders from outside AI agents: what does DoorDash’s pilot tell us?
The collected report describes an early DoorDash pilot for outside AI agents. Keep availability and measurement periods in view, then examine the path to a completed order. Company-wide order volume is not a suitable denominator for a pilot adoption rate.
What to consider
Start with who can actually use the integration and what they can do, then examine orders. The collected report quotes DoorDash CEO Tony Xu saying outside AI agents have generated only hundreds of orders since launch, while the company handled nearly one billion orders in the third quarter of 2026. It also describes an MCP connector less than a month old, available to a small set of early beta partners and mainly intended for business meal ordering. These are different populations and measurement periods; dividing the figures would not establish adoption or prove that demand is absent.
Read the pilot’s availability alongside the order count
Hundreds of orders describes the reported scale, but the summary gives no count of connected businesses, people reached or ordering opportunities per partner. Without that context, it cannot distinguish limited distribution from workflow friction or a lack of interest in ordering through an agent.
Retain the distinction between company-wide quarterly orders and orders accumulated since the connector launched. For your own pilot, define a measurement window, eligible users and tasks before comparing changes within that scope. An early integration is not a measurement of the entire market for agent commerce.
Separate intent to order from an order that actually exists
Record discovery, request initiation, item selection, confirmation and a valid order receipt separately. Retain the failure stage and any human handoff. This can reveal where the task stops instead of treating an agent’s claim of completion as a transaction.
For business meal ordering, choose an existing use case and ask the same trial group to try the established workflow and the new entry point. Compare completion, corrections and actual spending. This is a proposed evaluation, not DoorDash’s disclosed study design. The connector was not tested here, and no success rate is supplied.
Evaluate the in-house feature and outside connector separately
The summary also reports basket-value and item-variety results for Ask DoorDash, without a complete comparison population or method available here. It is a different entry point from the outside MCP pilot. Its results do not establish that the connector is mature or predict the basket value of another business.
Use the report as a prompt to consider availability, transaction outcomes and maintenance together. Before expanding a pilot, explain which existing problem it solves and how a failed attempt returns to ordinary ordering. At low volume, investigate missing stages; the summary does not support a return-on-investment or market-size estimate.
Source links
The links below were supplied with the collected records. This article offers independent analysis of their summaries; we have not individually checked the original sources or tested the products.