Does one failed inventory AI project say much about AI overall?

Mostly, it shows that a tool can fail at the everyday details that matter to a business. Inventory systems need to identify items reliably in real store conditions, not just perform well in a demo. Without more information about what went wrong, I wouldn't use one reported rollout to judge every kind of AI.

I'd separate two questions: did this system work well enough for the job, and are AI companies being valued sensibly? A disappointing operational project may be relevant to the first, but it doesn't prove the second is a scam. You'd need much broader evidence to make that case.

The useful takeaway is to ask what happens when the system gets things wrong. For inventory, a human check, a way to correct errors, and clear measures of accuracy should be part of the plan. If those pieces aren't working, replacing the tool can make more sense than keeping it just because it uses AI.