The tech industry is experiencing a peculiar mix of euphoria and turmoil. On one hand, revenues are at historic highs; on the other, layoffs have returned with a vengeance. In the first five months of 2026 alone, more than 115,000 tech workers lost their jobs—nearly as many as in all of 2025. And the explanation most commonly offered by executives? Artificial intelligence.

But according to Box founder Aaron Levie, the real culprit is something else: a collective delusion afflicting top leadership. He calls it 'AI psychosis.'

The CEO’s Dangerous Gap

Levie, a prominent AI optimist and active angel investor, took to X to articulate the problem. '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 wrote. In his view, executives play with AI tools, see a prototype generate a contract or a snippet of code, and then jump to the conclusion that AI agents can independently execute the entire workflow.

The gap is between what AI can do in a demo and what it takes to make it reliable in production. CEOs are not the ones debugging hallucinated library calls, training models on idiosyncratic contract terms, or spending days reviewing edge cases. They see the happy path—and assume the rest will just magically be handled.

This disconnect is not new. During the cloud computing boom of the early 2010s, many executives underestimated the complexity of migrating legacy systems, leading to runaway costs. The difference today is the speed and scale of layoffs tied directly to AI pronouncements.

Layoffs in the Name of AI

Several high-profile CEOs have publicly linked job cuts to AI adoption. Zeb Evans, CEO of project management startup ClickUp, proudly announced on X that he had laid off 22% of his workforce after deploying about 3,000 AI agents to handle internal tasks. Evans insisted the move wasn’t about cost reduction but about creating a '100x org' where humans oversee agents rather than execute work themselves.

This rhetoric echoes the 'productivity paradox' that has puzzled economists for decades. A meta-analysis published in October 2025 in UC Berkeley’s California Management Review found 'no robust relationship between AI adoption and aggregate productivity gain.' Research from the National Bureau of Economic Research did identify gains, but called them a 'productivity paradox' where perceived improvements are larger than measured ones.

Meanwhile, MIT researchers tracking LLM capabilities concluded that AI agents are still far from human-quality work. Their model suggests that by 2029, AI might achieve 80–95% success rates on text tasks, but true outperformance of humans is years away. In other words, the current wave of layoffs may be premature.

Historical Parallels

The tech industry has seen similar manias before. During the late 1990s dot-com bubble, executives slashed costs and fired staff on the assumption that the internet would render entire business functions obsolete. Many of those companies crashed when the promised efficiencies failed to materialize. The cloud era saw a similar pattern: early adopters overprovisioned and overspent, only to later realize that real automation required extensive configuration and change management.

AI psychosis might be the latest iteration. The difference is that today’s technology is genuinely transformative, but the hype has outrun the reality. A Harvard Business Review study warned that when everyone uses AI to produce more output, the bottleneck shifts to executive decision-making—people who must authorize and approve the flood of work. Without proper governance, chaos ensues.

The Voices of Reason

Not all tech leaders are succumbing to the delusion. Aaron Levie himself advises CEOs to 'use AI a ton' to truly understand its limitations. 'Come out the other side with an appreciation for both the upside and the real work,' he says. Other executives have publicly pushed back against the rush to automate human roles. They argue that AI should augment workers, not replace them, and that the most successful companies will be those that invest in retraining and upskilling.

The data supports caution. Even the most optimistic studies show that AI, in its current state, is best suited for narrowly scoped tasks under human supervision. Broad claims of AI-driven productivity gains remain unproven. And yet, the pressure on CEOs to demonstrate forward thinking often overrides empirical evidence.

In the meantime, the human cost continues to mount. Over 115,000 tech professionals have been laid off in 2026, with many citing AI as the reason. The irony is that while CEOs dream of fully autonomous organizations, the actual adoption of AI remains patchy and unstandardized. Companies that cut too deeply may find themselves unable to maintain the very systems they expect AI to run.

What’s needed is a sober reassessment of what AI can and cannot do. The technology holds enormous promise, but it is not yet the silver bullet that some executives imagine. Until CEOs close the gap between the demo and the deployment, the industry will continue to oscillate between euphoria and disappointment—with employees bearing the brunt of the psychosis.

Source: TechCrunch News