we stopped scoring AI skills on a scale and started classifying personas instead

building aisa.to — an AI that assesses peoples AI skills through conversation.

early on we scored everything on a 1-10 scale. seemed obvious. but the scores were kinda useless for actually helping people improve.

someone scoring 6/10 overall could be a completely different type of AI user than another 6/10. one might be great at prompting but never verifies output. another might be super critical of AI but barely uses any tools. same score, totally different problems, totally different advice needed.

so we built a persona classification system instead. 10 distinct personas based on the shape of someones skill profile across 5 dimensions, not just the average.

the copy-paster uses AI every day but accepts everything at face value. the sceptic has amazing critical thinking but underuses AI tools. the conductor orchestrates multiple tools into workflows. same "intermediate" score range, completely different users.

the insight that changed everything: people dont improve by "getting better at AI" generically. they improve by understanding their specific pattern and working on the gap thats actually holding them back.

turns out the persona approach also made the product way easier to explain to employers. "your candidate is a Copy-Paster" communicates more in 4 words than a spreadsheet of dimension scores ever did.

still iterating on this but its been the single biggest product decision we made. anyone else building assessment or classification products run into similar issues with scoring scales?