Google Says AI Now Generates 75% of Its New Internal Code

• Google said AI now generates 75% of its new internal code, with human engineers reviewing the output. 
• That figure rose from about 25% in October 2024 and 50% last fall, showing rapid adoption of AI coding tools inside the company. 
• Google is also using Cloud Next 2026 to center its enterprise strategy on AI agents, governance, and production deployment.

Google said 75% of its new internal code is now generated by AI and then reviewed by human engineers, marking a sharp increase from prior milestones and underscoring how deeply AI tools are now embedded in its software workflow. The update was disclosed on Wednesday as Google opened Cloud Next 2026 in Las Vegas. 

That number has climbed quickly. Google previously said AI generated about 25% of new code in October 2024, then 50% by late 2025. The move to 75% shows the company is no longer treating AI as a limited productivity tool. It is making AI central to how engineering work gets done. 

Google Shifts From AI Assistance to Agentic Workflows

According to sources, Sundar Pichai said Google is moving toward what he called “truly agentic workflows,” where engineers work alongside more autonomous AI systems. He said one recent complex code migration was completed six times faster than it would have been a year earlier, using engineers alone. 

Google is not only using AI to autocomplete or suggest code. It is increasingly deploying AI agents to handle broader engineering tasks inside larger workflows.

Additionally, the company is pushing staff to use AI across coding and other functions. That internal push has steadily raised the share of AI-generated code over time. 

Cloud Next Puts Enterprise Monetization at The Center

At Cloud Next 2026, Google said AI agents are becoming a core part of how it plans to monetize AI through business customers. The company introduced Gemini Enterprise, a broader umbrella for enterprise AI offerings that includes a reworked Vertex AI platform. 

Google also announced new governance and security features for AI agents, signaling that it wants to position its tools as production-ready infrastructure rather than experimental software. According to Google Cloud CEO Thomas Kurian, enterprise usage has shifted from traditional machine learning toward a rapid rise in customers building custom AI agents. 

Google is not only using AI heavily inside its own business. It is also presenting internal adoption as proof that agent-based tools are ready for larger commercial use. 

Numbers are Rising as The Industry Race Accelerates

Google’s update lands in the middle of a broader industry race around AI-generated software. Business reporting on the announcement noted that Microsoft, Meta, and Snap have all been increasing internal AI coding targets, though Google’s 75% figure is among the clearest and most aggressive disclosures so far. 

However, Google is framing the contest slightly differently. While some rivals are emphasizing coding assistants directly, Google used Cloud Next to highlight agents, enterprise deployment, governance, and infrastructure. It is saving some coding-related announcements for its I/O event in May. 

Google says AI is writing most of its new code, human engineers are reviewing and directing that output, and the company is building its next phase of growth around the same agent-driven model it is now using internally.

Related: Google Says Quantum Computers Could Break Crypto Sooner Than Expected