The Effects of Generative AI on High-Skilled Work: Evidence from Three Field Experiments with Software Developers
AI coding assistants raised completed tasks by 26% across three field experiments with 4,867 developers, with bigger gains for less experienced developers.
Summary
RCTs at Microsoft, Accenture and a Fortune 100 company randomised access to a code-completion tool.
Editor's opinion
Rigorous RCTs, but a developer-productivity tool-adoption study isn't a firm-strategy insight.
Key concepts
generative AI · software developers · productivity · field experiments