When the Boss Only Trusts the Bot: 'AI Psychosis' Is a Management Failure, Not a Tech One

According to the account making the rounds, an employee says their boss has started running every workplace choice past Claude and effectively ignoring human input, to the point of jibing that he suffers from "AI psychosis." The framing is anecdotal and clearly tongue-in-cheek, and there's no clinical claim here worth treating as one. We repeat it as reported, not as fact about anyone's mental state.

Strip away the punchline, though, and the story captures a genuine transition-era failure mode. AI assistants are good enough now to feel authoritative on almost anything, and that fluency is seductive: a chatbot always answers, never pushes back with office politics, and produces confident prose on demand. The temptation is to outsource not just the drafting but the judgment — and that's where a productivity tool curdles into an abdication of responsibility.

The real issue isn't the model; it's the org chart. A manager who "only listens to Claude" is telling his team their expertise doesn't count, and no language model — however capable — has accountability, context on the people involved, or skin in the game when a call goes wrong. Over-reliance like this erodes trust, hollows out the human relationships that make teams work, and concentrates decision-making in a system that was never designed to own outcomes.

Our reading: this is exactly the kind of short-term friction we should expect and name honestly. The long arc of these tools points somewhere genuinely good — freeing people from drudgery so they can spend judgment where it matters most. But that upside only materializes if we keep AI in the passenger seat: a brilliant advisor to interrogate, not an oracle to obey. The healthiest workplaces won't be the ones that adopt AI fastest, but the ones that stay clear about who is still responsible for the decision. Treat the model as a colleague you argue with, not a boss you submit to.