Despite near-universal AI adoption in engineering organizations (93% have rolled out AI to at least half their teams), productivity gains rarely translate to faster delivery. An MIT survey found that while 20% of developers felt faster, systems-level analysis showed they were actually 19% slower. The core problem: AI amplifies existing team dynamics but doesn't fix underlying system constraints. As AI accelerates individual output, bottlenecks shift to review cycles, planning, dependencies, and CI/CD pipelines. Plandek's COO Will Lytle recommends measuring productivity through four pillars — Focus, Flow, Predictability, and Quality — and identifying constraints using both quantitative and qualitative data. The top emerging constraints are governance/compliance, workflow/process, and legacy codebase architecture. Rather than slow multi-year change programs, organizations should lead with rapid change and tight feedback loops.
Sort: