There’s a certain wildness to the technology industry these days, with both things that mimic previous eras of big change like cloud computing (huge initial costs) and things we’ve never seen before (record profits accompanied by mass layoffs).
One possible explanation is that technology company executives, especially CEOs, collectively suffer from delusions of AI’s greatness. And at least one tech CEO is saying as much out loud. I’m Aaron Levie, founder of Box.
“CEOs are uniquely susceptible to AI psychosis because they are sufficiently removed from the last-mile work needed to create the most value with AI,” Levie wrote in X.
CEOs use Levie’s example to “play with AI” to develop prototypes and write contracts, then take the leap of faith that agents can do the job.
But these top-level executives are not the people who have to review code, find bugs, and identify calls to hallucinatory libraries before the software is deployed. As Levie points out, they are not responsible for training an AI model based on a company’s idiosyncratic contract terms, nor do they have to spend days poring over contracts to find sneaky terms.
In other words, Levy’s theory argues that CEOs don’t really understand processes well enough to know what can and cannot be automated. However, lack of knowledge does not prevent them from acting on their beliefs.
It’s important to note that Levie is not AI-hating. Quite the opposite. He mostly posts AI positivity about X to his 2.7 million followers, and writes a blog titled “Headless Software is the Future” about how software built for AI agents will move forward. He is also generous with his funds, supporting AI startups as an active angel investor.
So what should CEOs do instead? Levie advises CEOs to use AI “a lot” to really see what it can and cannot do, and “come out on the other side recognizing both the positives and the real work.”
I believe in humanity enough that I believe there are CEOs out there who are trying to do just that, but they seem to be in the minority at this point.
In the first five months of 2026 alone, there were almost as many layoffs in the tech industry as there were in all of 2025. So far in 2026, 152 tech companies have laid off 115,430 people, while 275 companies have laid off 124,636 people in 2025, according to industry layoffs tracker Layoffs.fyi.
And the majority of companies cite AI as the reason for these job cuts. Many argue that the largest tech companies are singling out AI, or crediting it with past or future productivity gains, when in fact other business decisions and metrics are driving the reductions.
Still, some of these stories are surprising. Zev Evans, CEO of project management and productivity software startup ClickUp, proudly declared in X that he had laid off almost a quarter (22%) of his employees after deploying around 3,000 AI agents to his internal operations.
Mr Evans asserted that this was not done to cut costs. Instead, they want a workforce made up of people who spend their days running AI agents and quickly reviewing their work. He believes this will create what he calls a “100x organization.”
Although AI is a very useful tool, data on AI and productivity do not support such assumptions. In miles.
A meta-analysis of other studies published in October in the University of California, Berkeley’s California Management Review found that “there is no robust relationship between AI adoption and total productivity gains.”
A study published in March by the National Bureau of Economic Research concluded that the introduction of AI has increased productivity, but noted the “productivity paradox” (perceived productivity gains are greater than measured productivity gains).
After creating thousands of agents to tackle tasks, MIT researchers came to the conclusion that in many cases, agents are not yet performing human-level tasks. They predict that at current rates of LLM improvement, models will be able to “complete most text-related tasks with an average success rate of 80% to 95% by 2029, at a minimally sufficient level of quality.”
In other words, AI is on track to have basic capabilities in most tasks within about three years. Researchers believe it will take several more years for agents to outperform humans.
Meanwhile, a study published in the Harvard Business Review showed that when everyone is using AI to produce more, the bottleneck simply shifts to the C-suite. Their work awaits those who must approve everything that everyone is creating. When everyone is empowered to act, OpenAI’s experience last year shows that things can get out of control.
Are CEOs ready for that? If not, the most certain outcome of the ongoing CEO AI psychosis will simply be organizational chaos.
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
