How should a tech org measure engineers when AI does much of the output?

Options

• Outcomes tied to business metrics — revenue, retention, incidents — not artifacts.
• Systems built: reusable platforms, factories, and paved roads others ship on.
• Lines of code and commit counts are dead; stop pretending they still correlate.
• Quality of judgment: what they refuse to automate, what they review carefully.
• Speed of learning and iteration with acceptable risk, not raw throughput.
• Team leverage — how much safer and faster the org becomes because they exist.
• Customer-visible reliability and time-to-recover when things break at 3 a.m.
• Documentation and teaching that reduces bus factor, especially for AI workflows.
• Ethical guardrails: data handling, security reviews, saying no to reckless prompts.
• Honest postmortems without blame theater — that culture is rarer than any model.