Google CEO Sundar Pichai says AI now writes 75% of new code at Google as software development enters a new era

Artificial intelligence is taking a far bigger role inside Google than many expected. Google CEO Sundar Pichai has revealed that AI now generates 75 percent of all new code created at the company, a sharp rise from 50 percent just six months ago. The announcement signals how rapidly AI tools are reshaping modern software engineering and how deeply Google is integrating them into its internal operations.

Pichai shared the update in a blog post during Google’s annual Cloud Next 2026 conference in Las Vegas, where the company outlined its latest enterprise AI strategy. The figure is one of the clearest indicators yet that large technology companies are moving from AI experimentation to full scale operational use.

According to Pichai, Google has been using AI to assist coding work internally for some time. However, the jump from half of all new code last fall to three quarters today shows how quickly confidence in these systems has grown.

This does not mean engineers have been replaced. Pichai made it clear that human developers remain responsible for reviewing and approving every line of code produced by AI systems. In practice, AI is being used as a fast drafting tool that can speed up repetitive work, generate solutions, and reduce time spent on routine programming tasks.

That balance between machine speed and human judgment is becoming central to the future of software creation. AI can produce output at scale, but trained engineers still decide what is reliable, secure, and ready for deployment.

The significance of Google’s 75 percent figure goes beyond one company. Google is among the world’s most advanced software organizations, managing products used by billions of people. If a company of that scale is trusting AI to generate most new code, it suggests AI coding systems have matured faster than many expected.

For the wider technology industry, the message is clear. AI is no longer limited to chatbots or content generation. It is now becoming a core engine of enterprise productivity.

Businesses around the world are likely to study Google’s model closely. If AI can safely accelerate engineering work at Google, many firms will attempt similar adoption strategies for internal development teams, product maintenance, testing, and modernization projects.

Despite growing automation, Google’s approach still places engineers at the center of the process. Pichai emphasized that every AI generated line is reviewed and approved by people.

This is an important distinction. High quality software demands accuracy, security checks, privacy safeguards, and long term maintainability. AI can suggest solutions quickly, but experienced engineers are needed to validate architecture decisions and catch hidden risks.

For developers concerned about job displacement, Google’s model currently points toward augmentation rather than replacement. Engineers may spend less time writing repetitive code and more time focusing on design, strategy, debugging, and higher value innovation.

Pichai also described a broader transition inside Google toward what he called agentic workflows. In this model, developers do more than request snippets of code from an assistant. Instead, they coordinate multiple AI agents that can handle tasks autonomously.

These digital task forces may be assigned responsibilities such as code migration, testing, bug analysis, documentation, or product prototyping. Engineers then supervise and direct the overall process.

Google says one especially complex code migration project completed through collaboration between agents and engineers was finished six times faster than similar work done by engineers alone a year earlier.

That claim highlights how the next stage of AI may focus less on simple assistance and more on coordinated autonomous systems capable of completing multi step workflows.

Another example shared by Pichai involved rapid prototyping. Google teams reportedly used an internal platform called Antigravity to move from concept to a fully functional native Swift app for the Gemini macOS release in just a few days.

If accurate, that pace would represent a major shift in product development cycles. Traditional app creation often takes weeks or months depending on scope, testing needs, and engineering resources.

Shorter development timelines can help companies respond faster to market demand, test ideas more quickly, and launch updates at a pace previously difficult to sustain.

Pichai said Google follows a customer zero strategy, meaning it uses its own products internally before selling them to outside clients. In this case, Google’s engineering organization acts as the first real world testing ground for the Gemini Enterprise Agent Platform.

The strategy allows Google to stress test tools under demanding conditions before offering them to global cloud customers. If systems can function at Google scale, the company believes they will be better prepared for enterprise deployment elsewhere.

This approach may also strengthen trust among business customers who want proof that Google relies on the same AI infrastructure it is marketing.

For software professionals, the shift suggests coding roles are evolving quickly. Strong programming fundamentals still matter, but new skills are becoming increasingly valuable. These include prompt design, AI workflow management, architecture review, security oversight, and integrating AI generated code into production systems.

For businesses, the opportunity is productivity. Faster development cycles can reduce costs, improve speed to market, and help teams modernize legacy systems more efficiently.

However, rapid adoption also carries responsibility. Companies must ensure proper governance, data protection, compliance, and human accountability when AI tools are deeply embedded into engineering workflows.

Google’s announcement may be remembered as a milestone in the rise of AI powered engineering. When one of the world’s largest technology companies says AI now creates most of its new code, it marks more than an internal statistic. It reflects a structural change in how software is built.

The future of coding increasingly appears to be collaborative, with humans setting direction and AI accelerating execution. Google’s latest disclosure suggests that future is arriving faster than expected.