Google: 75% of Code Generated by AI, a Major Turning Point

Google: 75% of Code Generated by AI, a Major Turning Point

At the Cloud Next 2026 event in Las Vegas, Sundar Pichai, CEO of Google, announced a striking statistic: 75% of the new code produced by the company is now generated by artificial intelligence. This figure, which was 50% just six months ago, reflects a rapid and massive adoption of AI in Google's development processes. This swift progression raises questions about the future pace of this trend and its implications for engineers.

Essential Human Oversight

Sundar Pichai's statement is clear: 75% of all new code at Google is generated by AI, but every line is approved by engineers. This means that AI does not operate entirely autonomously and does not have access to the Git repository. Engineers remain responsible for validation, highlighting the importance of their role in this automated process.

Google, a Laboratory for Its Own Technology

Google uses its own teams as a testing ground for its Gemini Enterprise platform before offering it to clients. This strategy, known as "Customer Zero," allows the company to test and refine its technologies internally before commercializing them on a large scale. This demonstrates how committed Google is to integrating AI into its daily operations. The speed of this adoption resembles less a mere experiment and more a full-scale deployment.

Towards Agentic Workflows

Beyond the impressive figure, the evolution towards "agentic workflows" represents a paradigm shift. Engineers are now orchestrating autonomous agents, which has enabled, for example, a complex code migration to be completed six times faster than before. Additionally, the Gemini application for macOS was developed in just a few days thanks to an internal platform called "Antigravity." This shift from chatbot to autonomous worker is not unique to Google, but the company may be the first to implement it at this scale.

Implications for the Future

This transformation raises questions about the future workforce at Google. If 75% of the code is already generated by machines, what will be the role of engineers moving forward? Sundar Pichai speaks of a vision of augmented productivity, but does not specify how this will affect long-term staffing needs. The structural question posed by this figure is how far this automation can go: 90%? 95%? And how many engineers will be needed to oversee these autonomous agents that work without pause?