Generative AI is increasing software development productivity. However, the gains are concentrated among experienced developers, according to a new study by the Complexity Science Hub (CSH).
The research analyzed large-scale software development data across six countries and found that the share of AI-generated code rose from 5% in 2022 to nearly 30% by the end of 2024. Over the same period, developer productivity increased by approximately 4%. Given that U.S. firms alone spend more than $600 billion annually on programming labor, even modest productivity gains carry significant economic implications.
However, the study reveals a clear divide by experience level. While early-career developers use generative AI more frequently—at an estimated rate of 37%—measurable productivity improvements are observed almost exclusively among senior developers. Researchers attribute this gap to differences in how developers interpret, validate, and integrate AI-generated code into broader systems. The CSH team found that experienced developers not only boost output with generative AI but also expand their technical scope. As a result, senior developers combine unfamiliar libraries and explore new domains, enabling deeper experimentation.
Industry leaders caution that productivity alone is an incomplete measure of AI success. Executives interviewed alongside the study emphasized that structure, accountability, and disciplined execution are essential for scaling AI beyond experimentation. Without these controls, AI adoption risks stalling or creating hidden quality issues within the software development lifecycle.
The findings also point to shifting career dynamics. As AI handles routine coding tasks, experienced developers gain leverage by focusing on architecture, edge cases, and system-level judgment. In contrast, less-experienced developers may struggle to extract similar value without guidance and oversight.
The study concludes that generative AI amplifies existing skill differences rather than flattening them. Organizations seeking sustained gains must pair AI tooling with training, review processes, and clear accountability to ensure productivity improvements translate into long-term value.
ソース：
https://www.zdnet.com/article/why-gen-ai-boosts-productivity-some-developers-not-others/
