Google CEO Sundar Pichai says 75% of the company’s code is AI-generated

Pichai wrote that Google is “staying on the cutting edge” by being “customer zero” of its own products. For example, he said, a recent “complex code migration” completed by both agents and engineers was done six times faster than what was possible just a year ago with engineers alone.

While the work has evolved, the fundamentals of Google’s engineering workforce have stayed the same, Google Cloud senior director and chief evangelist Richard Seroter told Fast Company.

With AI-generated code approved by humans—which Seroter called “critical in this era”—engineers are able to focus on “higher-value tasks like system architecture, design, and solving complex problems.”

At Google, the title of “software engineer” seems to have grown outdated. “Software engineers are becoming product engineers, or architects, as they move away from manual coding and toward an agentic operating model,” Seroter told Fast Company.

“Excitedly, many prior limits have dissolved,” he added. “No longer are Google engineers constrained by time or human energy, but rather can use AI to explore a seemingly endless array of ideas that benefit our users.”

It’s not just engineers utilizing AI tools at Google, though. According to Pichai’s blog post, Google’s marketing teams used AI models to “rapidly generate thousands of variations” of creative assets, which otherwise would have taken weeks.

“Using AI led to 70% faster turnaround and a 20% increase in conversions, getting us to market faster and more effectively,” Pichai said.

It’s a busy time for Google. At its Cloud Next 2026 conference, the company announced the launch of two new AI chips, as well as the release of a new Gemini Enterprise Agent Platform. At the conference, Pichai also said that Google will invest up to $185 billion on infrastructure to power autonomous AI agents. Google Cloud recently signed a multibillion-dollar deal with ex-OpenAI executive Mira Murati’s Thinking Machines Lab to expand on its AI infrastructure. Over the next few years, Seroter said that Google will be “prioritizing agent-first experiences.”

“The experimental phase of simple copilots is over. Tab completion, context-unaware chatbots, and ‘AI, please start this for me’ is no longer sufficient. We’re in the era of making AI and agents complete relevant work, steered by human operators,” Seroter said.

“For Google, the next few years are about transitioning from simple code generation to managed agency—where we provide a governed, enterprise-ready harness to build and scale autonomous agents,” he added.

Google I/O, the company’s annual developer conference, is scheduled for May 19, where Pichai said further announcements are in store.

When it comes to building and scaling AI agents, it doesn’t seem like Google will slow down anytime soon.

I agree. Over the last decade or so, regular search engines have become utterly useless for anything but buying crap and getting directions (to buy crap in person).

I always blamed it on SEO but after getting very good results with the AI tab this year, I am beginning to wonder if they made their standard search engine useless on purpose so that people like me would use AI. A few years ago I would test AI with searches that I already had answers for (like part numbers for engine components) and it would always fail. Always. Now it is quite good. (Like comparing AI generated images of hands three years ago to today... same for automotive stuff.)

And so, I am finding AI so useful now that I am no longer slamming the laptop shut in frustration after searching for information. It still has a lot of hallucinations, so you have to verify everything, but with AI now, I think I am getting more useful information more quickly than back before I felt that SEO had ruined search engines. Unfortunately, this makes me dependent on AI. Maybe by design...

Complete BS. He’s claiming that to avoid claims of using foreign labor.

"AI" means "Another Indian".

Of course, the Indian programmers are telling their managers that the code is generated by artificial intelligence systems. It is what the managers want to hear and report to the executives.

I have worked with Indian software teams before. They will always tell you what they think you want to hear. It seems to be a deeply embedded behavior that is required in their culture for survival and advancement. They have great difficulty overriding that behavior.

Accept those AI claims at your own risk.

They've moved it upstream a stop.

Cory Doctoroow's book Enshittification explains how the powers that be messed up saerch on purpose, so you'd spend more time and clicks on the search page, rather than leaving after getting the correct information immediately.

Thus exposing you to more ads and more money for Goolag from their business customers (if it's free, YOU are the product).

SO now they've done the same thing, in the AI space -- AI will give you "not quite" good enough answers in many cases, or code you in circles, so you have to pay more for tokens and they can (futilely) attempt to make up for the enormous spend (literally hundreds of billions) on physical plant for AI (not counting ongoing electricity and water costs, btw).

It will fail because, short of AI pr0n, there is not enough demand for AI at scale to justify sunk costs.

I’ve heard of that book (and of course have also heard that term used throughout the internet). Interesting points. I’ve long thought the “like, share, and subscribe” culture was going to be the next worst thing for the internet since SEO. There was to much shit to comb through already, now there’s clickbait to further waste my time? Enshittification for sure.

I’m a capitalist, but the emphasis on pure profit instead of focusing first on selling quality products is frustrating. I suspected the tech companies were purposely making their search engines less than optimal for searches (other than to directly sell you something) but did not put two and two together that they did this to keep you on the site to sell more advertisements. Makes sense.

I’m not sure how it will all end. I first noticed the SEO effects on ebay when searching for classic car parts. Used to scour ebay weekly for parts, then rather suddenly, no matter how I searched, I’d get pages and pages of car covers, wiper blades, and floor mats instead of very specific year and model fitment parts. Instead of buying parts weekly, I have now used ebay only a handful of times in the last six or seven years.

I pretty much gave up on regular search engines until they force fed me AI and it turned out to be useful. So that’s why I thought, “Ah, this is why they enshittified the internet.” But your/Cory’s explanation makes more sense.

Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.