The day ClickUp announced it was letting go of 22% of its workforce, the official line was terse: "AI agents will take over those roles." But Box co-founder Aaron Levie had already given this trend a name — 'AI psychosis,' a state in which an organization becomes so captivated by AI's possibilities that it loses sight of how the work actually gets done. What makes the condition more dangerous is that the people making the decisions are usually the ones who have never done the work themselves.

In 2026, the Reason for Layoffs Changed

As of the first half of 2026, tech-industry layoffs are already approaching 2025's full-year total. And it's not just the pace that has accelerated — the rationale has changed. In previous layoff cycles, the standard justifications were economic slowdown, over-hiring, and cost discipline. The layoff memos of 2026 feature different vocabulary: "AI agents," "automation transition," "role redefinition."

ClickUp's announcement is the archetype. For a project-management SaaS company, 22% is no small number relative to its size. The company's logic is that AI agents can now handle the repetitive work its people used to do. The premise isn't entirely wrong. Routine ticket handling, log triage, internal documentation cleanup — AI can absorb a substantial share of that work.

The problem is where that judgment was made. Figuring out which tasks are genuinely routine — and which only look routine — requires asking the people who actually perform them. And the people in the room deciding on layoffs have, more often than not, never done that work.

What AI Is Said to Do Versus What It Actually Does

The structure of what Levie calls 'AI psychosis' works like this: you watch a demo of AI doing something, and you believe it will perform the same way in a real work environment. Demos are built from controlled inputs, ideal conditions, and predictable outputs. Real work is not.

Consider a day in the life of a customer-support agent. If 60% of inquiries are classifiable, the other 40% involve emotions that defy categorization, judgment calls that depend on context, and situations that demand the company's unwritten institutional knowledge. AI handles the 60%. But the executive announcement simply reads, "AI handles customer inquiries." The 40% hasn't disappeared — the people have. Who handles that 40% isn't in the plan. Either the remaining team absorbs it, the customer experience quietly degrades, or a new job posting goes up a few months later.

There is, of course, a counterargument. AI transitions have genuinely strengthened some organizations. Certain companies handed repetitive work to AI and freed their remaining staff to focus on higher-value work. Salesforce and a number of enterprise companies have reported maintaining customer satisfaction while reducing overall labor costs after adopting AI. From this vantage point, AI-driven layoffs don't necessarily mean a loss of organizational capability. The claim that fast movers became more agile is hard to dismiss outright.

But those success stories share common conditions: people who understood the frontline work were deeply involved in designing the transition, and the scope handed to AI expanded gradually, through experiments. It was not fast decisions, big announcements, and across-the-board cuts. Before designing the solution, they first studied the perspective of the people who live with the problem every day. Whether you're building a product or transforming an organization, reversing that order changes the outcome.

Why This Pattern Keeps Repeating in Korean Organizations

When discussions about adopting AI begin at Korean companies, it's rare for frontline staff to raise the subject first. Direction gets set from above, a vendor demo moves up the chain, and a cost-savings report lands on leadership's desk. Along the way, the voice of the people who know the work best fades out. Tacit knowledge that can't be expressed in numbers never makes it into the report.

This structure persists not because anyone has bad intentions. Decision-makers see AI's potential and genuinely believe it's good for the organization. Frontline staff know how the transition will actually play out, but they often lack the language to make that case persuasively upward. The demo's logic is clean; the field's exceptions are messy. Reports choose the clean side.

If you're a solo operator or a director leading a small team, you face the AI adoption call yourself far more often. In that seat, the way to avoid 'AI psychosis' is to shrink the steps. Break down one team member's daily work first, run two-week experiments on which parts can actually be automated, and only then draw conclusions. Rolling it out across the whole organization comes after that.

If you're a middle manager, your most essential role when AI adoption comes up the chain is to document the frontline's actual work in concrete terms and carry it upward. The sentence "AI can handle 70% of this job, but the other 30% still needs a person — and here's why" has to make it into the report. Without that sentence, only the numbers remain.

Three things are worth checking. First, find out whether the person making the AI adoption decision has ever done the work in question. Second, if a wholesale transition is planned without a pilot period, question the premise. Third, confirm there's a plan for who handles the remaining work after the layoff announcement — and how. If none of these has a clear answer, the decision may well be mistaking a demo for reality.

In 2026, the moment for deciding whether to adopt AI has already passed. The question now is how to adopt it — and where to slow down. What ClickUp will only confront after letting 22% of its people go, an organization weighing the same decision today can see in advance. How well the person making that call understands the actual work will determine the organization's next year.