Tech industry leaders are increasingly facing a phenomenon described by Box founder Aaron Levie as AI psychosis. Levie argues that many chief executives suffer from delusions of grandeur regarding artificial intelligence because they are too distant from the operational details required to generate actual value. While CEOs often experiment with AI prototypes or automate high-level tasks, they frequently fail to understand the complexities of code review, bug detection, and the nuances of training models on specific company data. This disconnect leads to unrealistic expectations about what AI agents can currently achieve. Levie emphasizes that he is not an AI skeptic. He actively invests in AI startups and advocates for headless software designed for AI agents. His warning is that executives must deeply engage with the technology to appreciate both its potential and its limitations. Despite this advice, the current trend shows a minority of leaders taking such steps. Instead, many are driving mass layoffs under the banner of AI efficiency. Data indicates that the tech sector has already witnessed nearly as many job cuts in the first five months of 2026 as it did throughout all of 2025. With over 115,000 people fired across 152 companies, executives have cited AI as a primary reason for these reductions. Some leaders point to specific examples to justify these cuts. Zeb Evans, CEO of ClickUp, recently laid off 22% of his workforce after deploying roughly 3,000 AI agents. Evans claims the move was not about cost reduction but about transforming the organization into a high-performance unit where humans manage and verify AI output. However, empirical evidence does not yet support the notion of immediate, massive productivity gains. A meta-analysis in UC Berkeley's California Management Review found no robust link between AI adoption and overall productivity increases. Similarly, a study by the National Bureau of Economic Research highlighted a productivity paradox where perceived gains far exceed measured improvements. Research from MIT suggests that while large language models will likely reach a baseline competence level of 80% to 95% for text-based tasks by 2029, they are not yet capable of consistently producing human-quality work. Furthermore, a Harvard Business Review study indicates that widespread AI adoption shifts bottlenecks to executives who must approve the increased volume of output. If automation outpaces human oversight, the result could be organizational chaos. Experts warn that without a grounded understanding of AI capabilities, the current wave of AI-driven restructuring may lead to instability rather than the promised efficiency. The industry now faces a critical moment where distinguishing between hype and reality will determine long-term success.