The people deciding that AI can replace your job often haven't done that job in years - or ever. Box founder Aaron Levie calls this "AI psychosis": executives who are genuinely, confidently wrong about what AI agents can do today, because they're making decisions based on demos rather than direct experience.
The pattern is showing up in real numbers. ClickUp recently cut 22% of its workforce, with AI agents cited as the replacement plan. Tech layoffs in 2026 are already on pace to match all of 2025 totals. These decisions are accelerating fast.
The Gap Between Demo and Reality
The structural problem Levie identifies has a clear cause. A CEO sits through vendor presentations built to show best-case scenarios. The people actually doing the work - who encounter AI agent failures, hallucinations, and constant correction loops daily - rarely have meaningful input into those budget decisions.
This isn't a new dynamic. Enterprise software vendors oversold automation to executives in the 1990s and 2000s. What's different now is speed and irreversibility. When an ERP underdelivered, companies adjusted over years. When 22% of your workforce is gone and the AI agents produce unreliable output, there's no practical path back.
The productivity gains from AI tools are real but narrow. Repetitive, well-defined tasks - formatting, summarization, basic drafting, data extraction - see genuine efficiency improvements. Complex judgment calls, tasks requiring institutional memory, and work involving client relationships are harder to automate reliably. AI agents today still require regular supervision and human escalation paths.
ClickUp's Bet Will Set Industry Expectations
ClickUp's decision will be closely watched. The company has made a specific, public commitment: AI agents will handle work that 22% of its human employees were doing. If that holds, other companies will accelerate similar moves. If the agents underperform, ClickUp faces a real operational problem - and the industry gets an expensive case study.
Levie's point isn't that AI can't replace some jobs. It's that irreversible decisions about which jobs to cut are being made by people with the least direct experience of where the technology currently fails. That disconnect is where the psychosis lives - and right now, it's deciding a lot of people's careers.