How to measure AI developer productivity in 2025 with Nicole Forsgren
Nicole Forsgren (creator of DORA and SPACE frameworks, author of Accelerate, and now Senior Director of Developer Intelligence at Google Core Developer) joins Lenny's Podcast to discuss how teams should measure developer productivity in the AI era. She argues that "most productivity metrics are a lie" — especially lines of code, which AI makes trivially easy to game — and that traditional prescriptive metrics like DORA four must be used carefully because AI has changed feedback loops and review patterns. Her core thesis is that AI does accelerate coding, but developers aren't speeding up as much as headlines suggest because they still hit broken builds, flaky tests, and new bottlenecks around reviewing and trusting AI-generated code. She reframes the problem as Developer Experience (DevEx), which includes productivity, satisfaction, cognitive load, and flow state. She highlights trust as a new front-of-line concern: LLMs are non-deterministic, so engineers must now spend significant time evaluating output for hallucinations, reliability, and code style. She walks through her upcoming book Frictionless, which presents a seven-step process for setting up a DevEx program: start the journey with listening tours, secure quick wins, use data to optimize work, decide strategy and priority, sell the strategy, drive change at the right scope, and evaluate progress in a continuous loop. She also discusses how to measure AI's impact by aligning metrics with what leadership actually cares about — market share (focus on speed), profit margin (focus on cost savings), or transformation — and recommends starting with targeted surveys that ask developers to rank their top three friction points by frequency. On practical advice, she notes that AI is helping senior engineers build sophisticated multi-agent workflows that delegate parallel pieces of work, but emphasizes that engineers are still the architects. She recommends Claude Code as an underrated AI tool even for non-coding tasks, and discusses how AI is reshaping work blocks — potentially making even 45-minute sessions productive by helping engineers re-enter flow state.
Original sources
Podcast Lenny's Podcast
Episode:How to measure AI developer productivity in 2025 | Nicole Forsgren ↗
YouTube Lenny's Podcast
Video:How to measure AI developer productivity in 2025 | Nicole Forsgren ↗
AI Radar wrote this summary from the publisher’s notes with AI, not from a transcript. How it works.