Reuters
- A version of this story originally appeared in the BI Tech Memo newsletter.
- Sign up for the weekly BI Tech Memo newsletter here.
Charles Rollet has been digging into how tech companies are reorganizing around AI — especially Meta, which is leaning on AI coding tools to ship more product faster.
There are signs it's working: Meta ranks pretty high on revenue per employee, a metric Silicon Valley is increasingly focused on.
Here's my chat with Charles, edited for length and clarity:
You and other tech team colleagues have written a lot lately about how generative AI and AI coding tools are changing how tech companies operate. This is especially true inside Meta. What's the big takeaway from this research?
There's no more hiding from AI. Until recently, employees could get away with just being good coders or product managers. But now, it seems impossible to advance within Meta without embracing these tools — and touting your AI chops. We're already seeing these requirements formalized, with the company setting explicit goals for some engineers to produce 50%-80% of their code with AI assistance.
Why is Mark Zuckerberg pushing this so much at Meta? What's the goal for the company here?
There's a simple answer: productivity. Zuck wants 100x engineers who command armies of AI agents, rather than a bunch of junior devs doing the basics. That may be possible given recent progress. However, that also leads to a worrying trend if you're a Meta employee. If AI is so powerful, why should the company bother employing more than 76,000 very expensive employees?
Staff attending internal 'AI Weeks' might be digging their own graves, former Meta engineering director, Erik Meijer, recently told me. He worries that a business with as many users as Meta can't just crank out ten times more new features — its user base simply can't absorb that. So the logical alternative would be reducing headcount.
Which parts of Meta's business are being transformed the most by these "AI at Work" changes?
Reality Labs has had the most aggressive AI-based reorg within Meta, that we know of so far. The division's internal tools team, which is about 1,000 people strong, has radically reshaped itself, abolishing old job titles and centering work around small AI-native "pods." The team is rebranding employees as "AI builders" while managers will now be known as "AI pod leads" who use AI to help them do performance reviews.
News of that reorg spread to other parts of Meta, worrying some that the 'AI pods' model could be adopted more widely and be used as a justification for layoffs. Meta said in a reorg memo that headcount wouldn't be affected, but this speaks to the anxieties of some employees within the company.
Will this stuff work, or do you think there will be problems?
I think there will be a lot of problems, like an AI agent almost accidentally deleting your inbox, but they are mostly manageable. Claude Code seems to represent a genuine step change in AI capabilities, and to Meta's credit, they are not stopping employees from experimenting with all kinds of tools. For now, I'm more worried about startups vibe coding serious issues into their products than a better-run corporation like Meta.
Is this mostly about efficiency and cost savings, or is there something else going on here?
There are definitely other factors. A big push seems to be encouraging employees to be more flexible. For example, as part of the AI pods reorg, Meta told staff it expects engineers to do design work if that's what is needed to get things done. This could lead to some interesting products coming out of Meta, if roles aren't too strictly defined and people are empowered to build whatever they want. I do think it's a lot about efficiency and productivity, though. I wouldn't risk bragging about building an AI agent that does all of my work!
Sign up for BI's Tech Memo newsletter here. Reach out to me via email at abarr@businessinsider.com.
American Dollar Exchange Rate