We've assessed 105 candidates for AI skills — here are the insights

Over the past few months, we've run 105 AI skills assessments through AISA. Not multiple-choice tests. Not self-assessments. Real conversations where candidates demonstrate how they actually work with AI.

Here's what the data tells us:

Most people overestimate their AI skills.
The biggest gap we see isn't in whether people use AI — most do. It's in how they use it. Accepting the first output. Never iterating. Copy-pasting without verifying. Using one tool for everything when three better options exist.

Knowing tools ≠ knowing how to use them.
Candidates who can name every AI tool on the market often score lower than those who use two tools deeply. Vocabulary without practice consistently scores in the bottom half.

Critical thinking is the rarest skill.
Of our five assessment dimensions, Critical Thinking — the ability to evaluate AI output, catch errors, and understand where AI fails — has the lowest average scores. Most candidates trust AI output at face value.

The top 15% do something different.
They don't just use AI. They've built workflows around it. They iterate with intention. They know when not to use AI. And they can explain their reasoning — not just their results.

"My team uses AI" ≠ personal proficiency.
One of the most common patterns: candidates describe their organisation's AI strategy fluently but can't demonstrate personal, hands-on skill. Strategy awareness and personal capability are two different things.

We built AISA because we believe AI skills are the new literacy — and like any literacy, they need to be measured properly. Not with checkboxes. With evidence.

If you're hiring for roles where AI competency matters, the gap between what candidates claim and what they can demonstrate might surprise you.

aisa.to