Teams keep accountability when AI agents break production

Short summary

This essay argues that when AI agents implement code from specs, accountability remains with the team, not the agent or any individual developer. The author draws a parallel to CI/CD pipelines: automation moved execution but not responsibility. The real opportunity is moving review earlier to spec sessions where mistakes are cheap to fix, and the real challenge is that organizations have always pretended accountability was individual when it was always collective.

• •Team accountability doesn't change when AI agents write code — the team was always the unit of responsibility, not the individual
• •Spec review before agent implementation catches mistakes cheaper than code review after
• •Organizations must stop pretending accountability is individual; AI agents just make this tension harder to ignore

Generated with AI, which can make mistakes.