Engineering lead runs AI agent teams to ship security fixes and handle an incident in parallel

Over four days, an engineering manager used AI agents organized as a team of leads, coordinators, task workers and reviewers to handle security remediation, a production incident and its reviews, and a platform backlog at once. Roughly seventeen security fixes were merged and deployed behind feature flags, with the incident contained and its post-incident reviews written. The approach separated workstreams so each had its own owner, avoiding the context mixing that a single long assistant conversation produced. The manager's role centered on writing briefs, reviewing changes, and approving every production action. The article also covers guardrails, failures encountered, and the challenge of getting the wider team to adopt the model.