This article is a deep-dive from JudyAI Lab — an AI engineering playbook series with 100+ published guides, 5,000+ weekly readers across 60+ countries, focused on the practical side of running AI agents, trading systems, and content pipelines in production.

📰 Key Takeaways

Box founder Aaron Levie recently called out a phenomenon he calls “AI psychosis” - the executives who confidently decide “AI can replace this role” are often the people who understand that job the least. He points out this decision-making blind spot is spreading across tech, as leaders get so swept up in AI’s potential that they rush into large-scale layoffs without really understanding the workflows, responsibilities, and human judgment the job requires.

As a concrete example, collaboration tool platform ClickUp recently announced it was cutting 22% of its workforce, stating outright that AI agents would take over those roles. This wave of layoffs has pushed the overall scale of 2026 tech layoffs to nearly match the full-year total for 2025 - and it’s only mid-year, which shows AI-driven headcount reduction is accelerating.

Levie’s core argument isn’t against using AI - it’s a warning that companies bringing in AI to replace people often lack a deep understanding of what the work actually involves, which can lead to shortsighted decisions. When decision-makers aren’t familiar with the real complexity of the roles being replaced, they tend to overestimate how much AI can actually cover, which ultimately hurts the organization. This kind of “over-AI’d” thinking is quietly becoming a new source of management risk in Silicon Valley. Check the original link for the full interview.

💬 JudyAI Lab Take

The “AI psychosis” Aaron Levie is calling out is worth paying attention to for anyone thinking about deploying AI in the real world: the executives most eager to say “AI can replace this role” are often the ones who know the least about what that role actually does - and that’s the real blind spot in the decision.

ClickUp’s 22% layoffs, with AI agents stepping in to cover those functions, are part of why 2026’s tech layoffs have already nearly caught up to all of 2025’s total by mid-year. This trend should be a wake-up call for AI builders: the part of automation design that’s most likely to fail isn’t the tech stack you pick - it’s how deeply you actually understand the job you’re automating. When we design agent workflows without doing the deep interviews with the people actually doing the work, it’s easy to overestimate what AI can cover - you end up automating the visible steps while missing all the tacit human judgment and exception-handling underneath. What Levie is really pointing at is a requirements-analysis problem: the less you understand a job’s true complexity, the more AI you throw at it, and the more damage you might end up doing to the organization.

Before you plan any AI replacement, go talk directly to the people doing that job and ask “what’s the stuff you know that nobody would know unless you told them?” The answer is usually exactly the part AI struggles hardest to cover.

📅 Source Info

References

🔗 Further Reading