Past the Productivity Ceiling: Rebuilding the Enterprise from First Principles

AI can help your teams ship code faster than ever. But what if faster is exactly the trap? What happens when a company optimizes a workflow that shouldn't exist in the first place? GitLab's first CIO, Manu Narayan, argues that most enterprise AI rollouts are chasing the wrong scoreboard—lines of code, tickets closed, hours saved—and heading toward a productivity ceiling they can't engineer past. His answer: stop retrofitting AI onto pre-AI workflows and start rebuilding them from first principles, backed by a real operating model for governance, cost, and experimentation. Manu breaks down GitLab's hub-and-spoke model, why the team measures against business KPIs instead of token usage, how the human-in-the-loop role is shifting toward orchestrating agents, and why context and traceability are the differentiators that matter now. Watch or listen to the full episode: https://hatchworks.com/talking-ai/ -- Key Moments: 0:00 — Why a faster version of a pre-AI workflow is still a pre-AI workflow 4:30 — The hub-and-spoke operating model and AI transformation owners 9:06 — Skills vs. agents — human-invoked versus autonomous 16:13 — Why GitLab doesn't ascribe to token maxing 18:02 — Measuring adoption by role, beyond lines of code 27:39 — How human in the loop is evolving 30:49 — The manager-of-agents shift 33:52 — Why context and traceability are the new differentiators 37:29 — The bottleneck that moved from writing code to reviewing it 45:29 — The first move Monday morning: executive alignment -- Key Links: GitLab: https://about.gitlab.com Connect with Manu on LinkedIn: https://www.linkedin.com/in/manunarayan/