AI-Induced Delusions of Grandeur

Tech CEOs are suffering from AI psychosis, a phenomenon where they believe AI can perform tasks beyond its capabilities.

The Last Mile of Work

CEOs are uniquely prone to AI psychosis because they're distant from the last mile of work that generates most value with AI. They see the happy path results, not considering the next 10 or 20 things that have to happen. As Aaron Levie, founder of Box, said, "CEOs play with AI, develop a prototype, or generate a contract, and then make the leap to believing agents can do the work."

Lack of Knowledge

CEOs don't really understand processes well enough to know what really can and can't be automated. Levie advises CEOs to use AI "a ton" to see what it can and can't do. He advocates coming out the other side with an appreciation for both the upside and the real work.

The Consequences

In the first five months of 2026, the tech industry has already had nearly as many layoffs as in all of 2025. Many companies have pointed to AI as a reason for cutting jobs. Some CEOs believe AI can create a "100x org," where people review AI agents' work. However, research shows that AI adoption doesn't support such assumptions. A meta analysis found "no robust relationship between AI adoption and aggregate productivity gain."

The Future of AI

Researchers at MIT predict that AI will perform at base competence on most tasks in about three years. They believe agents will need another few years to outperform humans. Meanwhile, research published in the Harvard Business Review showed that when everyone is using AI to produce more stuff, the bottleneck shifts to executives. Their work awaits the people that must authorize all the stuff everyone is producing.

Are CEOs Ready?

If not, the most certain outcome of the ongoing CEO AI psychosis will be organizational chaos. The tech industry needs to address this issue before it's too late.