ClickUp’s 22 percent cut turns AI agents into a management-understanding test

In TechCrunch’s new video discussion about what happens when companies become too “AI-pilled,” the sharpest diagnosis comes from Box founder Aaron Levie: the people deciding that AI can replace someone’s job are often the ones least likely to understand what that job actually involves. TechCrunch frames that pattern through Levie’s phrase “AI psychosis,” a state where the real capabilities of AI systems get mixed with management’s desire to cut labor costs before it is clear what is being removed.

This is not an abstract productivity debate. In the supplied context, TechCrunch says ClickUp recently cut 22 percent of its workforce for AI agents, while tech layoffs in 2026 are already nearly matching all of 2025. Put together, those details change the tone of the story: AI is no longer just a tool entering existing teams, but a justification for removing teams before the system has proven it can carry the same operational load.

Box is a useful backdrop for that critique because the warning comes from enterprise software, where work is rarely a single clean task that can be replaced by a prompt. Inside companies, labor often lives in invisible transitions: who knows the exception, who understands why a process runs in that order, who reads the customer correctly, who can tell when an automated answer looks right but misses the situation. That is exactly the layer most likely to vanish when AI is introduced as a financial shortcut.

ClickUp is an especially pointed example because its products sit directly in the world of work organization, tasks and automation. The official ClickUp AI page points toward a product direction where AI can write, summarize and manage workflows. That is not the problem by itself. The problem begins when a productive tool is treated as proof that the entire human chain is replaceable, including judgment, prioritization, accountability and experience with edge cases.

That is why “AI psychosis” is abrasive but useful. It does not describe AI as irrational; it describes a corporate state in which a demo signal is mistaken for production evidence. An AI agent can handle parts of a process, reduce manual work and accelerate routine decisions. But there is a real gap between “this can help” and “this can take over the job,” and executives often skip that gap because the savings are visible immediately while the organizational damage arrives later.

The broader tech layoff pattern makes the warning harder to dismiss. If cuts are happening faster than companies can measure the long-term performance of AI agents, this is not simply optimization. It is an experiment on corporate operating memory. Customers, remaining employees and future products become the test surface for a claim that institutional knowledge can be replaced by a system that still has to prove reliability in real-world mess.

The sensible conclusion is not anti-AI. It is anti-magic. AI agents can be useful, but when they become a rationale for cutting people whose work leadership never fully understood, technology stops being a tool and becomes an alibi. That is the point where a clean efficiency story starts looking like a rushed reorganization with better marketing.