DoorDash's AI agents wrote code cleanups for $4.79 each. DoorDash estimates hours by hand.
InfoQ reported on 18 September how DoorDash automated the removal of stale feature flags, the on/off switches left in code after experiments.
DoorDash considers a flag stale when nobody has modified it for 90 days and the code still references it. Its experimentation platform manages more than 60,000 flags across about 623 repositories, with about 2,300 added each month. More than 1,000 were stale.
An orchestrator running on Claude Sonnet gathers each flag's history. An engineer approves every job before an agent touches the code. Claude Opus agents then edit the code in an isolated Git worktree. They run tests and static analysis before opening a pull request.
On a test set of 50 stale flags, 45 came back as usable pull requests. Each took 13.8 minutes and $4.79 on average. DoorDash estimated one to two hours for the same work by hand. 31 merged on the first pass. Another 14 needed revisions. The remaining 5 needed an engineer to step in. InfoQ reports no bugs or regressions.
Which ERP clean-up job keeps slipping because it is a thousand small fixes nobody has time for?
Sources
For your next steering committee
- 1Ask which AI tools already touch your ERP or finance data, including the ones nobody approved.
- 2Ask what each tool is allowed to change, and who reviews what it changed.
- 3Ask how the result is measured, and who decides to switch a tool off.
Issue 01 of the ERP LEADERS brief puts a timeline, the three-number review and six questions on one page. Read issue 01
Welcome back. · Issue 01