Google's AI Code Push

Google says artificial intelligence now generates more than 75% of its new code, with human engineers reviewing and approving the output. (blog.google) Chief Executive Sundar Pichai disclosed the figure on April 22 at Google Cloud Next in Las Vegas. He wrote that the company is “shifting to truly agentic workflows,” where software tools do more of the drafting and employees supervise the result. (blog.google) At the same event, Google Cloud announced a $750 million fund for its partner network to speed up agentic artificial intelligence projects for joint customers. Google said the money will support development, adoption, and training across a partner ecosystem it says includes 120,000 members. (googlecloudpresscorner.com) Agentic artificial intelligence refers to software systems that can take a goal, plan steps, use tools, and complete multi-step work with limited human input. Google used Cloud Next to pitch those systems as the next layer on top of chatbots and coding assistants. (cloud.google.com) The code figure turns Google’s internal engineering practice into a public benchmark. It also gives the company a simple number to show customers and investors that it is using the same tools it is selling through Gemini and Google Cloud. (businessinsider.com) The partner fund pushes the same strategy outside Google’s walls. Cloud resellers, consultants, and systems integrators often decide which models, tools, and infrastructure large companies adopt, and Google said the fund is meant to steer more of that work into its own stack. (crn.com) (googlecloudpresscorner.com) Google has been increasing that code percentage quickly. Pichai said in October 2024 that more than 25% of new code at Google was being generated by artificial intelligence, and reports this week said the figure had risen to 50% last fall before reaching more than 75% now. (blog.google) (businessinsider.com) Cloud Next also bundled the fund with new products meant to keep enterprise customers inside Google’s platform, including a Gemini Enterprise Agent Platform and new infrastructure announcements around Tensor Processing Units, Google’s in-house artificial intelligence chips. (blog.google) (crn.com) Google is not the only company pushing artificial intelligence into software development, but its scale makes the number hard to ignore. The company is saying, in public and with a dollar figure attached, that writing code and deploying workplace agents are now part of the same sales pitch. (businessinsider.com) (infotechlead.com)