Google issues company-wide guidance on using AI to code

Back in April, Sundar Pichai said over 30% of code at Google is now generated by AI. The company today released guidance for all Google software engineers on how to best adopt AI for coding and their work.

A source familiar with the matter tells us that Google on Monday emailed all software engineers recommendations and best practices for using AI. 

This extensive document was created by fellow engineers, and takes into account what has worked so far at Google. The expectation is for everyone to use AI to boost their productivity. The hope is that this will lead to more rapid iteration and boost development.

The guidance covers using AI assistance for coding, as well as understanding what the current generation can and cannot do. Google wants engineers across the company to understand how to build AI-based systems. More specifically, the goal is to have tech leads and managers understand how to incorporate AI development in teams.

Advertisement – scroll for more content

That 30% stat — which increased by over 5% in six months — involves AI generating code and humans reviewing it before it’s accepted. Today’s guidance emphasizes “maintaining rigor” across code review, security, and maintenance. Google also wants employees to explore using AI outside of coding.

Today’s guidance is the most formal the company has been about using AI, with existing adoption already quite notable. Pichai said earlier this month that AI usage has resulted in an estimated 10% increase in engineering velocity, or the amount of software work that’s accomplished in a certain period.  

Externally, this follows Google releasing Gemini Code Assist and Gemini CLI, as well as Gemini in Android Studio, in the past year.

In a statement, Google confirmed that it issued this AI guidance today: 

“Our engineering teams are already regularly using our AI tools for coding, fixing bugs, and much more. These recommendations and best practices, developed by a group of software engineers at Google, are designed to help all SWEs across the company get the most out of AI in their day-to-day work.”