DORA metrics are lying to you and AI is making it worse

DORA metrics measure delivery throughput but not system comprehension, and AI-generated code is widening that gap. As AI tools accelerate deployment frequency and reduce change failure rates on paper, teams may lose genuine understanding of their critical systems. The author argues that DORA's blind spots include out-of-band changes from vendors, AI agents, and identity providers that never enter the traditional SDLC review process. Practical recommendations include adding detailed scope notes to DORA reviews, requiring incident narratives alongside metrics, and testing whether team members can explain critical paths end-to-end. The core message: DORA is a scoped instrument, not a complete health signal, and leaders must stop treating it as one.