The 'AI Psychosis' Risk: Why Tech CEOs May Be Overestimating Automation

The tech industry is currently grappling with a disconnect between executive optimism and operational reality. Box CEO Aaron Levie recently coined the term "AI psychosis" to describe a phenomenon where top-level leaders, insulated from the granular details of daily workflows, overestimate the current capabilities of artificial intelligence. According to Levie, because CEOs often interact with AI only through polished prototypes or simplified tasks, they fail to grasp the complex "last mile" of work—such as debugging code or verifying nuanced contract language—that remains essential for actual business value.

This disconnect carries significant real-world consequences, most notably in the form of aggressive workforce restructuring. In 2026 alone, tech companies have already neared the total layoff figures seen in 2025, with many executives citing AI-driven productivity as a primary justification. Some leaders, such as ClickUp CEO Zeb Evans, have openly replaced portions of their workforce with AI agents, aiming to transition toward a model where humans serve primarily as overseers of automated processes. This trend suggests that many firms are betting their operational future on a level of AI maturity that may not yet exist.

However, empirical data currently contradicts the narrative that AI is driving massive, immediate productivity gains. Academic research, including studies from UC Berkeley and the National Bureau of Economic Research, has yet to find a robust link between widespread AI adoption and aggregate productivity growth. By acting on the assumption that AI can seamlessly replace human expertise, executives risk damaging their organizational capacity and morale. Levie suggests that for leaders to avoid these pitfalls, they must move beyond high-level strategy and engage deeply with the technology to understand both its genuine potential and its current limitations.