Box CEO Coins 'AI Psychosis' to Explain Tech Exec Overconfidence on Automation
Summary
- • Box CEO Aaron Levie argues tech executives suffer 'AI psychosis' — overestimating AI readiness by relying on polished demos
- • The theory emerges as 2026 tech layoffs approach full-year 2025 totals, with most companies citing AI as justification
- • Academic research from Berkeley, NBER, and MIT finds perceived AI productivity gains consistently exceed measured outcomes
- • Critics warn AI-driven layoff announcements may be 'AI washing' — cover for standard cost-cutting decisions
Details
Box CEO Aaron Levie coined 'AI psychosis' to describe exec overestimation of AI readiness
Levie argues CEOs are uniquely susceptible because they interact with polished AI demos and prototypes but are insulated from the downstream work — code review, bug discovery, hallucinated library calls, contract training — that reveals AI's real limitations. His theory: distance from execution creates distorted confidence.
Levie is an AI booster, not a skeptic — 2.7M X followers, active AI angel investor
His credibility as a critic comes precisely from being an AI advocate who backs AI startups and promotes AI-first software design. He advises CEOs to use AI heavily to develop a realistic picture of both its upside and its real limitations.
2026 tech layoffs through May: 115,430 people from 152 companies, near full-year 2025 total
By comparison, 124,636 people were laid off across 275 companies in all of 2025, per Layoffs.fyi. The pace of 2026 cuts is sharply accelerating, with a majority of companies citing AI productivity gains as rationale for headcount reductions.
ClickUp CEO Zeb Evans cut 22% of staff after deploying ~3,000 AI agents for a '100x org'
Evans claimed the layoffs were not cost-driven but a strategic shift toward a workforce of people who run and rapidly review AI agents. Critics cite this as a textbook example of AI psychosis — or AI washing of cost-cutting decisions.
Multiple studies find no robust link between AI adoption and aggregate productivity gains
A UC Berkeley California Management Review meta-analysis found 'no robust relationship between AI adoption and aggregate productivity gain.' An NBER study identified a productivity paradox where perceived gains outpace measured ones. MIT researchers concluded AI agents aren't yet doing human-quality work but forecast 80–95% of text tasks handled by 2027.
Critics say AI-driven layoffs are 'AI washing' — cover for business decisions driven by other factors
The concern is that executives attribute headcount reductions to AI efficiency gains when the underlying drivers are standard cost pressures or strategic pivots. AI becomes a socially acceptable rationale that deflects scrutiny and signals modernity to investors.
Insight = attributed analysis and argument, Context = background framing, Industry Update = business and workforce developments, Research = academic or empirical study findings
What This Means
The 'AI psychosis' framing captures a real and consequential gap: executives who shape hiring, investment, and strategy are making high-stakes decisions based on AI capabilities they experience through demos rather than deployment. If the research consensus holds — that perceived productivity gains exceed measured ones — the current wave of AI-justified layoffs may be creating organizational fragility without the promised efficiency. For AI practitioners, this dynamic underscores the importance of surfacing last-mile implementation realities to leadership before automation decisions are made.