The tech industry is experiencing a peculiar kind of madness. In the first five months of 2026 alone, more than 115,000 workers have been laid off, nearly matching the toll of the entire previous year. Executives frequently point to artificial intelligence as the catalyst. Yet Box founder and CEO Aaron Levie believes many top-level leaders are suffering from what he calls 'AI psychosis' — a collective delusion about what AI can realistically achieve.
Levie, a prominent angel investor in AI startups and an outspoken advocate for AI-driven transformation, posted his diagnosis on X: '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 argues that CEOs play with prototypes, generate a contract, or demo a chatbot and immediately assume agents can replace the messy, detailed work of code review, bug fixing, and contract analysis that still requires human judgment.
This disconnect is not new. During the early cloud computing boom, companies burned through capital on runaway infrastructure costs before learning proper governance. But the current wave combines record revenues with unprecedented layoffs. The scale is staggering: according to Layoffs.fyi, 115,430 employees have been let go from 152 tech firms in 2026, compared to 124,636 from 275 companies in all of 2025. Many of those cuts are explicitly linked to AI, leading critics to accuse firms of 'AI washing' — attributing layoffs to AI when financial or strategic reasons are the real drivers.
One notable case is ClickUp, the project management startup. CEO Zeb Evans proudly announced on X that he had laid off 22% of his workforce after deploying approximately 3,000 AI agents to handle internal tasks. Evans insists the move was not about saving money; he envisions a '100x org' where humans supervise agents rather than do the work themselves. Yet his enthusiasm contradicts available data.
A meta-analysis published in October 2025 by UC Berkeley's California Management Review found 'no robust relationship between AI adoption and aggregate productivity gain.' Another study from the National Bureau of Economic Research (March 2026) did detect some productivity improvement but noted a 'productivity paradox' where perceived gains far outstrip measured ones. MIT researchers, analyzing the capabilities of AI agents, concluded that current models produce human-quality work only in narrow domains; they predict LLMs will handle most text tasks with 80–95% success by 2029, and that outperforming humans will take several more years.
Meanwhile, a Harvard Business Review article highlighted that if everyone uses AI to produce more output, the bottleneck simply shifts to managers who must approve everything. OpenAI experienced this firsthand: when the company empowered agents to act autonomously, chaos ensued. Levie's advice is to 'use AI a ton' — but do so with eyes wide open, appreciating both the upside and the real, unglamorous labor that remains.
The roots of this 'psychosis' may lie in the tech culture of visionary leaps. Executives are trained to see potential, not friction. They are rewarded for bold claims. But the gap between demo and deployment is vast. For example, training an AI model on a company's idiosyncratic contract terms requires weeks of labeling and validation, all overseen by humans. And even advanced code assistants like GitHub Copilot hallucinate libraries or produce insecure code.
If CEOs continue to act on their AI delusions, the result will not be a sleek automated future but organizational chaos, as HBR warns. Teams will be gutted, surviving employees will be overwhelmed reviewing agent output, and the promised productivity gains will remain out of reach. Levie's prescription is simple but rarely followed: spend enough time in the trenches to know what AI can and cannot do. Until then, the industry will keep firing people based on fantasy, not fact.
Source: TechCrunch News