Google announced Wednesday that around 75% of its new code is now generated by AI and later reviewed by engineers, a significant increase from a year ago. The company said engineers are transitioning from writing code directly to managing AI systems, with productivity improving rapidly over the past year.
The company disclosed in October 2024 that AI was responsible for about 25% of its code output. By late 2025, that share had reached 50%. The latest milestone signals a broader transition, with AI now taking the lead role in Google’s software development pipeline instead of acting mainly as an assistive layer.
The shift is largely driven by Google’s internal rollout of Gemini models, now used by engineers for code generation, refactoring, and migration tasks. The company has also broadened AI adoption across non-engineering teams, with usage in certain areas reportedly factored into performance assessments.
Sundar Pichai framed the development as part of a larger operational shift, with Google advancing toward what he described as “truly agentic workflows.” In practice, that model places engineers in supervisory roles over systems able to execute complex tasks with limited direct input.
Sundar Pichai highlighted tangible efficiency gains from the company’s new workflow model. He said a recent high-complexity code migration, handled jointly by AI agents and engineers, was completed six times faster than comparable engineer-only processes from a year earlier.
The emphasis on agentic systems marks a move beyond basic code-completion tools toward software that can independently plan and carry out development tasks. In practice, that includes codebase migrations, large-scale refactoring, and other work that previously required extensive human coordination.
The deployment of AI tools has not been consistent across Google. Some teams within Google DeepMind have been permitted to use Anthropic’s Claude Code alongside internal models, a decision that has reportedly led to friction over tool standardization and strategy.
Google’s approach is similar to what many big tech companies are doing, with AI-generated code becoming a normal and important part of development.
Microsoft reported similar developments, with Chief Executive Satya Nadella saying in April 2025 that 20% to 30% of the code in some projects was written by AI. Around the same period, Chief Technology Officer Kevin Scott said he believed 95% of code would be generated by AI within five years.
Meta has established formal targets for AI use within its engineering teams. Internal documents indicate that by the fourth quarter of 2025, the company aimed for 55% of code changes in certain groups to be agent-assisted. During the first half of 2026, some teams are expected to have 65% of engineers using AI to write more than 75% of their committed code.
At Snap, the shift is already visible in day-to-day operations. The company said earlier this month that at least 65% of its new code is now generated by AI under its current operating model.
The trend signals a structural change in software development, with engineers moving toward supervisory roles over automated systems rather than manually producing most code. At Google, where AI already accounts for the majority of new code output, that transition is already in progress.
Maybe you would like other interesting articles?