Google’s ‘Project EAT’ Is Its Big Push to Make Staff Use AI

According to Business Insider, Google has an internal initiative called “Project EAT” that aims to supercharge employees with AI and transform the company into an “AI-powered workplace.” The project was created in May 2025 within the AI and Infrastructure unit, known as AI2, which is led by senior VP Amin Vahdat. The effort started as a grassroots push and has a 12-week seed-stage, with internal tests already showing “promising signs of improved developer velocity.” Google’s leadership, including CEO Sundar Pichai, has explicitly warned employees that rivals are using AI and they must do the same to compete. The project’s name is a reference to “eating their own dog food,” meaning employees internally test products first.

The internal AI mandate is real

Here’s the thing: this isn’t just a friendly suggestion from the cafeteria. This is a top-down, competitive mandate. When Sundar Pichai and engineering VPs are sending company-wide emails telling you to use AI for coding, that’s a direct order. It’s fascinating that Project EAT started as a grassroots effort, but you can bet it got immediate executive sponsorship once leadership saw it as a vehicle for this broader push. The internal mission statement reads like corporate Kool-Aid, promising “dramatically higher productivity” and “better work-life balance.” But let’s be real. The primary driver here is fear. Fear that Microsoft, with its deep OpenAI integration, is eating their lunch. Fear that developers elsewhere are coding 20% faster with Copilot. Google‘s response? Make everyone use the tools, and fast.

Dogfooding on a grand scale

So they’re “eating their own dog food.” That’s a classic tech strategy, and it makes sense. You can’t sell AI-powered workplaces to other businesses if your own house isn’t in order. Piloting everything within Amin Vahdat’s AI2 org is a smart move—that’s the group building the chips and data centers that power this whole AI race. If the tools work for them, they’ll probably work for anyone. But there‘s a huge risk here, too. What if the tools aren’t that good? Forcing adoption of subpar internal tools can actually *slow* people down. Engineers are famously skeptical. They’ll use what works best, whether it’s Google’s internal tool, GitHub Copilot, or some open-source model. The project’s goal to “mitigate risks associated with the rapidly evolving external AI landscape” sounds a lot like they want to lock people into the Google ecosystem. Good luck with that.

The productivity paradox

The internal FAQ cites “improved developer velocity, reduced toil, and enhanced code quality.” I’m skeptical. We’ve seen these promises with every new tech paradigm. Sometimes it pans out, often it doesn’t. Throwing AI at every process—from engineering to product management to operations—is a massive cultural and technical shift. It’s not just about giving people a chatbot. It’s about rethinking workflows, dealing with hallucinations in code suggestions, and managing a new layer of dependency. And let’s not forget the infrastructure cost. Vahdat’s org is swallowing billions in capex. Every query to an internal AI model costs real money in compute. Will the promised productivity gains actually offset that astronomical spend? That’s the billion-dollar question.

The bigger picture: Google vs. itself

Basically, Project EAT is a microcosm of Google’s entire AI challenge. They have the talent, the infrastructure, and the data. But can they move with the urgency and focus of a startup or a rival like Microsoft? This internal push is as much about changing Google’s own DNA as it is about building better tools. It’s about combating inertia. The risk isn’t just that the tools fail. It’s that employees, faced with a clumsy mandate, become cynical and disengaged. Turning a company of this scale into an “AI-powered workplace” is a moonshot. And as we’ve seen with other moonshots, sometimes you land among the stars, and sometimes you have a hard landing. For a company that relies on robust, reliable systems, betting the internal farm on rapidly evolving AI is a fascinating, high-stakes experiment. We’ll see if they’re building the future or just creating a new kind of corporate toil.