Let Them Eat Snacks
How one of tech’s strongest engineering cultures became collateral damage in the AI race
Meta has just given us the first great case study in how not to do an AI transformation. Over the last few months, the company has taken a butcher’s knife to what was once regarded as one of the strongest engineering cultures in tech.
It cut 10% of the workforce, created a new Applied AI org that many employees did not choose to join, and forcefully moved thousands into AI-related roles. Some workers described themselves as “draftees.” Others called the work “soul-crushing.” The smartest people I know at Meta are either gone, recently laid off, or still there with one foot out the door.
Andrew Bosworth has now admitted the rollout was “atrocious,” and Mark Zuckerberg acknowledged the company made mistakes. Employees are pushing back on the nature of the work, the lack of choice, the surveillance concerns, the loss of autonomy, and the feeling that the company was asking for extraordinary commitment while communicating that everyone was increasingly replaceable. That is a hard circle to square.
The mistake was not that Meta moved aggressively. In AI, speed matters. The mistake was assuming that speed required making employees feel disposable, disoriented, and conscripted.
AI adoption is not just a resource-allocation problem, it is also a collective-action problem. Every company in the world is trying to move people toward AI. The question is whether employees feel like they are being invited into the future or dragged into it.
You cannot lay people off, reassign thousands more into work they did not sign up for, stretch managers across absurd spans of control, monitor employee activity for training data, and then solve the resulting morale problem with snacks, offsites, and a hackathon. That is corporate Febreze.
One of the stranger management mistakes of the AI era is assuming that it makes people less important. If anything, the opposite is true.
When technology is stable, process matters. When technology is changing, talent matters.
In tech, the ability to attract and retain the best people has always been a leading indicator of company performance. You could have predicted that OpenAI and Anthropic would matter because great people wanted to work there. Anthropic’s advantage, in the current moment, is not just technical. It is cultural: a clearer sense of mission, purpose, and trust. The entire AI industry is a case study in talent concentration.
Meta seems to have missed the lesson hiding in plain sight.
A mature company can often survive mediocre culture because the machine already knows what to do. An AI transition is different. There is no machine yet. The company needs its best people to invent one, but is instead giving them every reason to leave before it exists. It is asking them to think like founders while treating them like furniture. That is a difficult strategy.
The winning move is not to aggressively “AI-ify” the workforce, but rather to handle the human layer of AI adoption with care. You cannot automate trust, purpose, or belief. At least not yet.