Silicon Valley has a new diagnosis - "AI psychosis." The term was put into circulation by Aaron Levie, founder of Box, but not to describe users going mad from chatbots, rather to describe the executives so swept up in artificial intelligence that they've completely lost touch with how it actually works on the ground.

Levie's idea is simple and uncomfortable for many at the top: bosses who make billion-dollar decisions based on presentations and slides, but have never touched the tool themselves, don't know what they're buying. They are, as he puts it, "removed from the last mile of the work" - precisely the part where artificial intelligence has to prove it's worth it, and most often fails.

The debate comes at a moment when resistance to AI is starting to show among ordinary users too, not just sceptical analysts. Students at graduation ceremonies boo every mention of artificial intelligence. Installations of the DuckDuckGo search engine jumped 30 percent, as people flee the forced feeding of AI results on the big search engines. When an AI can't even spell its own company's name correctly, it's hard to sell the story that this is the future.

Not everyone sees catastrophe. Part of the industry thinks the question isn't whether AI is useful, but whether companies know what they're offering. Those who get results, it seems, do so by focusing on a concrete need rather than scattering ten features at once. The difference between a tool that solves a problem and a tool that's looking for a problem to justify its own existence is exactly what "AI psychosis" misses.

For a Balkan reader, this isn't an exotic Californian topic. The same logic - buy the latest thing because everyone's talking about it, without checking whether you need it - we see everywhere, from state digital projects to firms paying for software nobody ever opens. Levie's question holds everywhere: how many of the decision-makers have actually tried the work they're signing off on?