Google reports halving code migration time with AI help

Google successfully employed AI-powered large language models (LLMs) to speed up internal code migrations, applying them to tasks like changing ID sizes, upgrading libraries, and replacing packages. This approach reduced migration time by 50-89% compared to manual methods. In one project, 80% of modifications were AI-generated, with 87% committed without changes. Google stresses the importance of human oversight and recommends using LLMs in conjunction with traditional migration techniques.