Box founder Aaron Levie just put a name to something a lot of workers have been feeling: “AI psychosis.” His argument, highlighted on TechCrunch AI’s Equity podcast, is sharp. The people deciding AI can replace your job are often the ones who understand your job the least. That gap between executive confidence and ground-level reality is becoming one of the defining tensions of 2026.
What stands out here is that both sides of the AI debate are right at the same time. The believers see real capability. The skeptics see real overreach. TechCrunch AI reports that hosts Kirsten Korosec, Anthony Ha, and Sean O’Kane dug into exactly this collision, and it’s worth paying attention to because the stakes just stopped being theoretical.
The numbers are getting loud
Three data points from the discussion tell the story:
- ClickUp cut 22% of its workforce, reportedly to lean into AI agents.
- Tech layoffs in 2026 are already nearly matching all of 2025, and we’re not even halfway through the year.
- DuckDuckGo installs are climbing as users flee Google’s AI-stuffed search results and just want links back.
That last one matters more than it looks. It’s a quiet signal that not everyone wants AI injected into every product they touch. When users start switching tools to escape your AI features, that’s the market talking back.
Why this matters now
There’s a difference between automating a task and replacing a role. AI is genuinely good at the first. The second requires understanding all the messy, undocumented work a person actually does, the judgment calls, the context, the relationships. Levie’s point is that the executives making the cut decisions rarely have that visibility.
This is significant because the cost of being wrong is asymmetric. Cut too aggressively based on a demo, and you discover the gaps after the institutional knowledge has walked out the door. Move too slowly, and a leaner competitor eats your margin. Neither error is cheap.
We’re watching companies pick a lane in real time. Some are using AI as a genuine productivity layer that makes existing teams faster. Others are using “AI” as cover for cuts they wanted to make anyway. The market hasn’t sorted out which approach wins yet, and the layoff numbers suggest a lot of bets are being placed on the aggressive side.
The skeptics aren’t Luddites
The DuckDuckGo migration reframes the whole debate. The pushback isn’t anti-technology. It’s anti-forced-AI. People want AI where it helps and absent where it doesn’t. Products that respect that line will hold users. Products that bolt AI onto everything because the board wants an AI story risk pushing customers toward simpler alternatives.
That’s a real strategic warning for any company shipping AI features right now. Adoption is a choice, not a default.
Practical takeaways
For businesses and AI practitioners trying to stay on the right side of this:
- Map the work before you cut the role. Talk to the people doing the job. The undocumented 30% is usually where the value hides.
- Treat AI as augmentation first. Make your current team faster before you assume AI can replace them outright. The data on reversed AI layoffs is already piling up.
- Give users an off switch. If your AI feature can’t be turned off, you’re betting every customer wants it. DuckDuckGo’s growth says some don’t.
- Watch the re-hiring signal. Companies quietly bringing back roles they automated are telling you where the capability line actually sits.
Levie has been a steady, credible voice in enterprise software for over a decade, and he’s not anti-AI. That’s what gives the “AI psychosis” framing weight. It’s coming from a builder, not a doubter.
The companies that win the next year won’t be the most AI-pilled or the most AI-skeptical. They’ll be the ones who figure out where AI genuinely replaces work and where it just looks like it does on a slide. Full breakdown, plus three deals and Waymo’s new robotaxi, is on TechCrunch AI’s Equity podcast.