What happens when companies become too AI-pilled?

The phrase “AI psychosis” is a sharp, uncomfortable diagnosis. It suggests a collective delusion—a detachment from the human fabric of work that becomes possible when you’re viewing the organization from the highest floor or the farthest abstraction. The person who sees a role as a series of tasks to be optimized is rarely the person who understands the contextual judgment, the relationship management, the creative problem-solving, or the sheer human stubbornness that often makes a job valuable. This isn’t a new corporate phenomenon; think of the classic tension between headquarters and the shop floor. But AI supercharges it. It offers a seductive, quantitative veneer: a dashboard that claims a task takes 3.2 minutes, so why not automate it? This ignores that the other 57 minutes of that hour might involve a critical decision, a reassuring client call, or a novel workaround that only a human who cares could devise.

The real psychosis, then, may be in the metrics themselves. When a company like ClickUp cuts 22% of its workforce, the immediate narrative is one of strategic agility—leaning into the future. But what is lost? Institutional memory evaporates. The nuanced understanding of why a certain process was clunky, or what customers truly complained about, walks out the door. You replace a person who could intuit a problem with a system that can only react to its parameters. The efficiency gain is real, but it’s narrow. It’s the efficiency of a mill that can grind wheat faster, without asking if anyone still wants to eat bread.

This is where the human cost transcends economics and becomes existential. The layoffs aren’t just numbers; they are people who built their expertise in a paradigm that is being actively discarded. They are being told their skills are obsolete, not because the outcomes they achieved are no longer needed, but because the method is being replaced by a cheaper, faster, more controllable alternative. The profound irony is that the very qualities that make humans resilient and creative—adaptability, empathy, deep contextual knowledge—are the first casualties of this psychosis. We are automating not just the labor, but the learning that labor represents.

The speed of the 2026 layoff wave is particularly telling. It’s no longer about individual companies over-hiring and correcting; it’s a sector-wide recalibration based on a shared bet about AI’s trajectory. This creates a homogenizing pressure. When every tech firm watches every other firm and decides, “Well, if they’re doing it, we must, too,” it ceases to be a rational business decision and becomes a panic—a psychosis of its own. The question shifts from “Is this right for our customers and our culture?” to “Are we falling behind in the efficiency race?”

Levie’s observation points to a crucial power imbalance. Those making the replacement decisions are often insulated from the operational reality. They see cost centers; employees see communities. They see inefficiencies; employees see necessary relationships. This disconnect is dangerous because it invites a future where work becomes infinitely scalable but utterly hollow. We risk building organizations that are algorithmically perfect but humanly barren, able to execute functions but incapable of generating the serendipity, trust, or passion that defines great work.

The path forward isn’t a Luddite rejection of AI. The technology holds genuine promise for eliminating drudgery. The challenge is to implement it with empathy, not psychosis. That means involving the people who do the work in the redesign of their jobs. It means measuring what truly matters—customer satisfaction, innovation, team resilience—not just the speed of a single output. It means having the courage to say that some jobs are more than their tasks, and that the future of work should augment human potential, not simply replace it at the behest of those who never fully understood it in the first place. The alternative is a deepening psychosis where we optimize ourselves out of meaning.