The Truth About Developer Productivity in the AI Age (IT'S A TRAP)

AI-driven hype is pushing organizations backward toward activity and output metrics (AI prompt counts, percentage of AI-generated code, pull request throughput) instead of outcome metrics that actually predict business success. Drawing on the DORA/Accelerate research, the author argues these AI usage metrics are low-value, easily gamed, and violate Goodhart's Law. A critique of a DX industry report illustrates how vendors conflate tool adoption with engineering strategy and correlation with causation. The recommended alternative: use Accelerate metrics (deployment frequency, lead time, change failure rate, time to restore) as leading indicators, ladder them up to business outcomes, and only then optionally track AI usage as a secondary activity metric with explicit caveats about its limited meaning.