Aaron Levie, Box CEO, told tech CEOs they have "AI psychosis": they prototype agents on the happy path and skip the last-mile work agents can't yet handle. 115,430 tech layoffs in five months of 2026 already match all of 2025 (124,636 across 275 companies). Most cite AI as the reason.
A CEO of a sleepy-sounding storage company just called out his peers. The line lands because the numbers behind it land harder.
Levie's argument: executives play with AI, generate a contract or write a script, and skip the next 20 steps - reviewing terms, wiring up legacy systems, training on company-specific edge cases. Then they cut headcount based on the prototype. 2026 tech layoffs are tracking ahead of all of 2025 already. Five months in. Most companies cite AI as the cause.
For PMs: when an exec says "agents can do this," ask what the last-mile work looks like. Make them write it down. For execs: Levie is publicly daring his peers to use AI more before they cut. Take the dare. For boards: pressure-test layoff plans against actual agent reliability data, not demos.
⚡ Why this matters
- First major SaaS CEO to name the disconnect between executive AI demos and the actual work agents can do.
- Pairs a viral framing ("AI psychosis") with concrete layoff numbers that make it harder to dismiss as just talk.
- Tells you the demo-to-deploy gap is now an executive-level problem, not just an engineering one.
🔍 What happened
- May 27, 2026. TechCrunch publishes Levie's X post and follows with a podcast interview.
- Fortune publishes a longer Levie piece on May 29 connecting AI psychosis to the layoff wave.
- Levie's claim: CEOs are "sufficiently distant from the last mile of work" to over-rotate on agent demos.
- Layoff data (cited by Fortune): 115,430 layoffs across 152 tech companies in Jan-May 2026.
- Comparison: 124,636 layoffs across 275 companies for all of 2025.
- Most affected companies cite AI as the primary driver for cuts.
💬 Smart takes
- Aaron Levie (Box CEO): "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's prescription: "Use AI a *ton.* Figure out the real implications of agents in the enterprise. Come out the other side with an appreciation for both the upside and the real work that goes into them."
- TechCrunch (May 31 follow-up): Making sense of the debate. The pattern is consistent across companies that cut deepest in 2026.
- Skeptic counterpoint: Some cuts are overdue trimming after 2021-2022 over-hiring, not pure AI overreach. The two signals are mixed.
🧭 Where this goes
- More public exec voices challenge the "AI replaces X% of headcount" narrative by end of Q3. Levie cracked the dam.
- Boards start requiring "agent reliability evidence" alongside layoff approval requests by FY27 planning.
- Government data (BLS, EU labor) starts breaking out AI-cited cuts as a separate category by year-end.
- Hiring rebounds in last-mile and validation roles - QA, contract review, customer success - as the limits of agents land.
- A class of "agent ops" jobs emerges to handle the work CEOs assumed was free.
🎯 Implication
- For PMs: when product asks "can an agent do this end-to-end," run the full happy-path-plus-edge-cases test. Document the gaps.
- For execs: separate the productivity story from the headcount story. Both can be true. Conflating them is the trap Levie names.
- For investors: watch which AI-rationalized layoff companies have to re-hire by Q4. That's the AI psychosis test.