In boardrooms, earnings calls, and tech conferences across Silicon Valley, a new trend has emerged among tech CEOs: proudly proclaiming how extensively they use their own artificial intelligence products internally.
Microsoft CEO Satya Nadella recently revealed that between 20-30% of the company's code repositories now contain AI-generated code. "I'd say maybe 20%, 30% of the code that is inside of our repos today and some of our projects are probably all written by software," Nadella stated during a conversation with Meta CEO Mark Zuckerberg at LlamaCon.
Not to be outdone, Google CEO Sundar Pichai announced that AI is generating "well over 30%" of the company's new code, up from 25% just six months ago. These executives are using these statistics to demonstrate their confidence in their own AI systems.
Meanwhile, Zuckerberg has set an even more ambitious target, predicting that "maybe half the development is going to be done by AI, as opposed to people" at Meta within the next year. He's also been vocal about developing AI that can function as "a sort of mid-level engineer" at the company.
Beyond code generation, tech leaders are embracing AI dogfooding in other ways. Uber executive Sachin Kansal has taken his dogfooding approach to the next level, completing 700 trips delivering food or people to test the company's services firsthand. He's now applying this same methodology to Uber's integration of autonomous vehicles, personally testing Waymo robotaxis in Austin.
This practice of "eating your own dogfood" isn't new in tech—it dates back to the 1970s and was popularized at Microsoft in the 1980s—but it has taken on new significance in the AI era. Companies are using it not just as a marketing tactic but as a practical approach to identify bugs, improve user experience, and demonstrate the safety and reliability of their AI systems.
However, as AI becomes more deeply integrated into critical business operations, the stakes of dogfooding are higher than ever. When tech CEOs publicly commit to using their own AI products extensively, they're making a statement about their confidence in these systems—and setting expectations for the rest of the industry to follow suit.