The rise of AI coding assistants such as GitHub, Copilot, Cursor, Claude Code, and Windsurf, along with agentic development workflows, is forcing engineering teams to rethink how they measure developer productivity. Traditional metrics such as lines of code, pull request volume, and commit counts were already imperfect.
Now, with AI generating significant portions of code, those measurements often reveal less about actual impact than they once did. To better understand what should replace them, we asked developers, engineering leaders, and AI practitioners to share their perspectives.
Focus on Outcomes Instead of Output
"AI can generate code incredibly fast, but writing code was never the goal. Solving customer problems is the goal. Productivity should be measured by outcomes rather than output." — Louis Leung, Co-Founder and Developer, inFlow Inventory
According to Leung, many traditional engineering metrics become less meaningful when AI can generate hundreds of lines of code within minutes. A developer producing fewer lines of code may actually be delivering greater business value if AI handles routine implementation work.
