There is a certain wildness in the tech industry these days that both mimics previous eras of large changes, like cloud computing, and is like nothing we’ve ever seen before. Record revenues are accompanied by mass layoffs, and the driving force behind this paradox appears to be what some are calling 'AI psychosis' — a term coined by Box founder and CEO Aaron Levie.
'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,' Levie wrote on X. According to his theory, executives play with AI, create a prototype, or generate a contract, and then leap to believing that AI agents can independently handle the entire workflow. But these top-level leaders aren’t the ones who review code, discover bugs, or identify hallucinated libraries before deployment. They are not responsible for training AI models on company-specific terms or spending days combing through contracts for sneaky clauses. In short, CEOs lack a deep understanding of the processes they seek to automate, yet that ignorance doesn’t stop them from acting on their beliefs.
The Scale of the Problem
The consequences of this mindset are stark. In just the first five months of 2026, the tech industry has seen nearly as many layoffs as in all of 2025: 115,430 people have been fired from 152 companies so far in 2026, compared to 124,636 people let go by 275 companies in 2025, according to Layoffs.fyi. The majority of these companies have cited AI as a reason for the cuts, though many argue that this is 'AI washing' — using AI as a convenient excuse for decisions driven by other business metrics.
One prominent example is Zeb Evans, CEO of project management software startup ClickUp. He proudly declared on X that he had laid off almost a quarter of his employees (22%) after rolling out about 3,000 AI agents to handle internal work. Evans insisted the move was not about cost reduction but about creating a '100x org' — a workforce composed of people who run AI agents and spend their days quickly reviewing the agents’ output. While such bold claims attract attention, the data on AI and productivity tells a different story.
What Research Says — or Doesn’t
A meta-analysis published in October in the UC Berkeley California Management Review found 'no robust relationship between AI adoption and aggregate productivity gain.' This conclusion echoes findings from the National Bureau of Economic Research, which acknowledged a 'productivity paradox, in which perceived productivity gains are larger than measured productivity gains.' In other words, executives feel they are getting more done, but the numbers don’t back that up.
At MIT, researchers created thousands of AI agents to perform various tasks and concluded that agents are simply not producing human-quality work yet. They predicted that, at the current rate of LLM improvement, models will 'be able to complete most text-related tasks with success rates of, on average, 80%–95% by 2029 at a minimally sufficient quality level.' That is still about three years away, and even then, the researchers believe agents will need another few years to outperform humans. For now, the gap between CEO expectations and AI reality is wide.
Organizational Bottlenecks and Chaos
Meanwhile, research published in the Harvard Business Review highlighted a critical bottleneck: when everyone uses AI to produce more output, the burden shifts to executives who must authorize all the new material. If everyone is empowered to act without oversight, organizations risk chaos — something OpenAI experienced last year when uncontrolled AI usage led to internal turmoil. Levie himself is not an AI skeptic; he is an active angel investor in AI startups and regularly posts about 'headless software' as the future. But his warning stands: CEOs need to use AI 'a ton' to truly understand its limits and potentials, and 'come out the other side with an appreciation for both the upside and the real work.'
The historical context is instructive. During the early cloud computing boom, companies rushed to migrate workloads, only to face runaway costs and security breaches. Similarly, the dot-com bubble saw massive layoffs after overinvestment in unproven technologies. Today’s AI hype cycle may follow a similar pattern, with CEO overconfidence leading to wasted resources and organizational disruption. The difference is the speed: AI agents can be deployed at scale in weeks, not years, magnifying both positive and negative effects.
Another factor is the pressure on CEOs to demonstrate innovation to investors and boards. In quarterly earnings calls, mentioning 'AI-driven efficiency' can boost stock prices even if the actual productivity gains are negligible. This creates a perverse incentive to announce layoffs linked to AI, even when the true drivers are cost-cutting or restructuring. The result is a cycle of hype and disappointment that harms employees and erodes trust in AI’s potential.
Levie’s critique is not an indictment of AI itself but of the leadership mindset that conflates experimentation with operational readiness. He advocates for a hands-on approach: CEOs should personally use AI tools for weeks, testing edge cases and failures, to calibrate their expectations. Without that, the most certain outcome of the ongoing CEO AI psychosis will simply be organizational chaos.
Source: TechCrunch News