the metric that keeps surprising us after hundreds of AI assessments
been sitting on this data for a while and it keeps coming back in conversations so figured id share.
we run AI skill assessments at aisa.to — conversational, not quiz-based. after hundreds of sessions the number that keeps surprising us is Safety: 45/100 average.
that means most people aren't thinking about data privacy, hallucination checking, or when NOT to use AI. the overall average is 52/100 with 62% scoring below Proficient.
the other surprise: high usage does NOT equal high skill. some daily ChatGPT users score lower than people who use AI once a week but verify everything.
this shifted how we build the product. we stopped optimising for "do you know AI tools" and started measuring what people actually do with them. full data: aisa.to/state-of-ai-fluency
Interesting. How are you measuring whether it is working?