the dimension nobody scores well on (from 412 AI assessments)
building aisa.to — a conversational AI that assesses people's AI skills through a 20-40 min chat.
after 412 assessments, Safety & Responsibility is consistently the lowest scoring dimension at 45/100. across all personas, all experience levels. people who are genuinely good at prompting and workflow integration still score terribly on things like checking for bias, understanding privacy implications, or having any verification process for AI output.
the uncomfortable part: the people who score highest on capability dimensions are often the ones with the biggest safety gaps. they're moving fast and shipping with AI but rarely stopping to ask whether they should.
we're building role-specific certification paths now because generic "AI skills" doesn't capture this nuance. a marketer using AI for copy has completely different safety considerations than an engineer using it for code generation.
412 assessments, 52/100 average, 62% below Proficient, and the safety gap is the one that keeps me up at night.
aisa.to/state-of-ai-fluency has the full breakdown.
The safety gap finding is the most interesting thing in this data — and the inversion is what makes it worth paying attention to. The most capable users having the biggest blind spots isn't surprising in hindsight, but it's the kind of thing you need 412 data points to say with confidence.
The role-specific certification angle is the right move. "AI skills" as a category is too broad to be actionable — a marketer and an engineer are making completely different risk tradeoffs.
Curious: when you're ready to take aisa.to beyond the assessment phase and into broader distribution, what does your launch plan look like? Communities, enterprise HR channels, something else entirely?