412 AI skill assessments later — the pattern nobody expected

we've run 412 conversational AI assessments through aisa.to and the one thing that keeps surprising us: the people who rate themselves highest almost never score highest.

the actual top performers talk about failures, limitations, and when they choose not to use AI. the ones who say "AI handles everything for me" usually can't describe a single process for checking whether the output is actually correct.

we built 11 criteria across 5 dimensions and 10 distinct user personas from this data. the finding that keeps coming back: confidence and competence are basically uncorrelated when it comes to AI skills.

working on a larger 1000-sample report now. curious what other founders have seen with AI adoption in their own products.