Aaron Levie warns of AI psychosis among tech CEOs

Aaron Levie, the founder of Box, argues that tech executives are suffering from “AI psychosis.” According to Levie, this condition occurs because leaders are too far removed from the actual execution of tasks to understand what can realistically be automated. Levie stated, “CEOs are uniquely prone to AI psychosis because they’re sufficiently distant from the last mile of work that still has to happen to generate most value with AI.” He notes that while executives may build a prototype or generate a contract using artificial intelligence, they do not see the daily work required to review code, fix bugs, or train models on specific company terms.

This executive mindset is increasingly linked to organizational restructuring and job cuts. In the first five months of 2026, 115,430 employees have been fired from 152 tech companies, according to Layoffs.fyi, an industry tracker for tech sector layoffs. This nearly matches the 124,636 employees let go by 275 companies in all of 2025. Many of these firms are accused of “AI washing”—a practice where companies credit AI for productivity gains when other business decisions are actually driving the results. However, some executives are explicit about the connection. Zeb Evans, the CEO of ClickUp, announced a 22% staff layoff after rolling out about 3,000 AI agents for internal work. Evans stated this was not to cut costs, but to build what he calls a “100x org,” which he defines as a workforce optimized for AI agent management.

Academic research, however, challenges the assumption that AI is ready to replace human labor at scale. A study published in October in UC Berkeley’s California Management Review found no robust relationship between AI adoption and aggregate productivity gain. Similarly, research published in March by the National Bureau of Economic Research concluded that while AI adoption improved productivity, it created a productivity paradox where perceived productivity gains are larger than measured productivity gains. Additionally, researchers at MIT predict that by 2029, AI models will be able to complete most text-related tasks with success rates of, on average, 80%–95% at a minimally sufficient quality level. This timeline suggests that AI is still years away from outperforming human workers on complex tasks.

Why it matters

Tech executives are increasingly making major organizational decisions, including mass layoffs, based on a potentially flawed belief in AI’s immediate productivity gains. If these perceived gains do not translate into actual business performance, companies risk facing organizational chaos.