Does your CEO have AI psychosis? Aaron Levie thinks most of them do.

The intersection of corporate efficiency and generative artificial intelligence has birthed a new organizational pathology: "AI psychosis." Coined by Box CEO Aaron Levie, the term describes a growing disconnect where C-suite executives, driven by the seductive promise of autonomous agents, are aggressively downsizing workforces based on a fundamental misunderstanding of what their employees actually do. While the allure of a leaner, high-margin organization is perennial, the current trend suggests that executive leadership is increasingly prone to overestimating the cognitive capabilities of current AI while underestimating the nuanced, tacit knowledge inherent in human labor.

This phenomenon does not exist in a vacuum. It follows a decade of Silicon Valley’s "growth at all costs" mantra, which has pivoted sharply toward "efficiency at all costs" in the wake of rising interest rates and cooling venture capital. Major tech players and startups alike have spent the last 18 months signaling to Wall Street that AI is the ultimate overhead-reduction tool. The context is a labor market already under immense pressure; tech layoffs in the early stages of 2026 are already pacing to rival the somber totals of 2025. What was once seen as a tool for augmentation is increasingly being marketed by SaaS vendors—and accepted by CEOs—as a wholesale replacement for middle management and creative roles.

The mechanics of this shift rely on the deployment of "AI agents"—software capable of executing multi-step tasks with minimal human oversight. Companies like ClickUp have pioneered this aggressive transition, recently shedding nearly a quarter of their workforce in favor of automated systems. From a business standpoint, the logic is seductive: agents don’t require benefits, they don't suffer from burnout, and they scale infinitely. However, Levie’s critique highlights a mechanical flaw in this reasoning. Most executive-level "job descriptions" are abstractions that fail to capture the problem-solving, cultural navigation, and edge-case management that human workers perform daily. When those humans are removed, the hidden "glue" holding complex workflows together often vanishes with them.

The industry implications of this "psychosis" are twofold. First, it risks a hollowing out of corporate institutional memory. If entry-level and mid-tier roles are automated away, the pipeline for future leadership dries up. Second, there is a mounting regulatory and social risk. As more firms trade human livelihoods for algorithmic efficiency, the push for "AI taxes" or stringent labor protections for displaced workers is likely to gain political momentum. Competitively, companies that succumb to AI psychosis may find themselves agile in the short term but brittle in the long term, unable to innovate once their human creative engine has been dismantled.

Market analysts are now beginning to question the sustainability of this "replace-first" strategy. While cutting 22% of a workforce may provide a temporary bump in stock price or runway, the technical debt of replacing nuanced human judgment with probabilistic models can be catastrophic. If an AI agent hallucinates a customer contract or mismanages a supply chain ripple, the lack of human oversight could lead to liabilities that far outweigh the savings in payroll. We are currently in the "honeymoon phase" of AI deployment, where the novelty of the tech masks the systemic risks of total reliance.

Moving forward, the industry must watch for the "re-entry" phase—the moment when companies that over-automated realize they have lost their competitive edge and begin a frantic search for the talent they recently discarded. We should also monitor the emergence of "Human-Centric" branding, where companies may eventually market their reliance on human intelligence as a premium feature or a mark of reliability. The test for the modern CEO is no longer whether they can integrate AI, but whether they have the wisdom to discern where the machine ends and the essential value of human ingenuity begins. The "psychosis" will only break when the real-world costs of automation failures finally hit the balance sheet.