Box CEO Aaron Levie Says Tech Executives Are Suffering From Mass AI Psychosis, and the Data Backs Him Up

Box CEO Aaron Levie has a diagnosis for the strange behavior of his peers: AI psychosis. In a late May post on X, Levie argued that tech executives are uniquely vulnerable to a distorted view of artificial intelligence because they are too far removed from the actual work required to make the technology function reliably.

CEOs are sufficiently distant from the last mile of work that still has to happen to generate most value with AI, Levie wrote. They play with the tech, see a prototype work once, and extrapolate that into a belief that autonomous agents can replace entire workflows. The people actually reviewing code, hunting down hallucinations, and building the scaffolding around these tools see a much messier picture.

That gap between executive perception and ground-level reality has become a defining tension of the current tech cycle. Levie, who runs a content management platform serving enterprise customers, is not some AI skeptic. He is describing what he sees as a cognitive trap that leads to overpromising and underdelivering.

The happy path Levie describes is the demo that works perfectly once. A CEO generates a contract, spins up a prototype, or watches an agent complete a task in a controlled setting. The leap from that single success to assuming agents can handle production workloads is where the psychosis sets in.

According to TechCrunch, Levie pointed out that these executives never have to deal with the next 10 or 20 things that must happen to get sustainable results. Code review, bug discovery, edge case handling, and hallucination detection all fall to workers who see the technology's limitations up close.

Fortune reports that this observation aligns with broader data. A 2025 survey cited in their coverage showed a widening gap between executive AI optimism and employee skepticism. Workers who use AI tools daily report higher rates of frustration with accuracy and reliability than their bosses acknowledge. The pattern is not new, but the stakes are higher when companies are betting their futures and their headcounts on AI's promise.

TechCrunch notes a peculiar contradiction defining this moment: record revenues accompanied by mass layoffs. The industry has seen boom cycles before, but the current combination of financial success and aggressive headcount reduction is historically unusual.

Futurism frames the dynamic more sharply, describing the job market as a barren wasteland and calling the AI investment frenzy one of the largest financial bubbles in recent memory. Executives are pursuing complete automation while workers face the consequences of those bets not yet paying off.

Levie's comments land in the middle of this tension. He is not arguing against AI investment, he is arguing against the magical thinking that leads CEOs to cut jobs before the technology can actually replace them. The psychosis is not about believing in AI, it is about believing the hard part is already done.

The disconnect is not just a management theory, it is showing up in therapy offices across the Bay Area. The San Francisco Standard reports that psychotherapists treating tech workers are seeing a surge in patients with existential despair tied directly to AI.

Candice Thompson, a Menlo Park therapist, told the Standard that roughly 80 percent of her patients work on or with AI. She dates the flood of AI-related anxiety to the past three to six months. Patients are not just burned out, they are grappling with apocalyptic fears about their professions and their sense of purpose.

Thompson described a shift in how she evaluates these concerns. In the past, a patient saying this is the end of the world would clearly be psychosis. Now, she said, a majority of those fears are things we actually have to consider. The line between clinical paranoia and rational assessment of technological disruption has blurred.

The Augmented Educator captured a moment that crystallized the mood. At UC Berkeley's main commencement, former Labor Secretary Robert Reich mentioned artificial intelligence exactly once. The stadium erupted in boos.

The publication's editor, who attended his son's PhD ceremony, recorded the reaction as a raw expression of generational anxiety. These graduates are entering a job market where AI is pitched as both savior and replacement, and they are not buying the happy path narrative.

That reaction stands in stark contrast to the boardroom enthusiasm Levie described. While CEOs see AI demos and imagine cost savings, graduates see a labor market that has not yet figured out how to integrate them alongside increasingly capable machines. The boos at Berkeley were not about the technology itself, they were about the story being told about it.

Levie's critique does not come with a neat policy prescription, but the implications are clear. Companies that cut too deep based on AI promises they cannot yet fulfill risk hollowing out the institutional knowledge needed to actually make the technology work.

Fortune notes that the data on AI productivity gains remains mixed. A Berkeley research review cited by Futurism identified seven persistent myths about AI and productivity, suggesting that the evidence for transformative gains is thinner than executive rhetoric implies.

The therapists quoted by the Standard are not seeing the crisis abate. If anything, the pace of layoffs combined with the hype cycle is accelerating the anxiety. Levie's term, AI psychosis, may stick because it names something many workers already feel: their bosses are living in a different reality about what AI can do today.

Whether the disconnect narrows depends on how quickly the last mile of work gets done, and whether CEOs are willing to acknowledge it exists at all.