1,219 AI skill assessments later — here's what the data actually says

been running aisa.to for a while now and just crossed 1,200 assessments. some patterns are becoming hard to ignore.

average score across all users: 48/100. the median is even lower. 62% of people score below what we'd call 'Proficient' — and these are self-selected people who chose to take an AI skills test, so presumably more interested in AI than average.

the weakest dimension? Safety, at 45/100. people are building workflows around AI tools without thinking about data exposure, hallucination risks, or when NOT to use AI. that gap is going to bite companies hard.

we just got accepted into UKAI (UK AI consortium) which feels like validation that this problem matters beyond just our little corner of the internet.

the counterintuitive thing: the people who score highest aren't the ones using the most tools. they're the ones with the most systematic approach to checking AI output. verification > volume, every time.

would love to hear from anyone else building in the AI skills/assessment space. what patterns are you seeing?