Your candidate gets a task from your own vacancy and works it out with an AI assistant, the way they would on a normal working day. You get a level from 1 to 5, with their own work as evidence.
Your first three candidates are free.
A. Vandenberghe
Assessment report
Builder
“Where does the 23% figure come from? It is not in anything I gave you, so take it out.”
Checked a number in the AI’s draft against the source and removed it.
Dokima turns it into a test with knowledge questions and a practical task from that job. It takes a few minutes.
A name and an email address are enough. Each candidate gets their own link and picks the moment.
A level from 1 to 5 per candidate, with a quote from their work behind every criterion. You can change any score yourself.
A marketer and a developer land on the same five levels, so you can put them side by side.
Knows what AI can and cannot do, and where it goes wrong.
Gives clear instructions, judges the output, picks the right tool.
Delegates work in several steps, manages context, checks the result.
Turns prompts into reusable building blocks, small tools and agents.
Designs agent systems, weighs risk, decides how a team uses AI.
Since February 2025, Article 4 of the AI Act asks organisations to take measures for the AI literacy of their people. It was amended on 27 July 2026: the measures have to fit your AI systems, roles and risks. There is no minimum level and no certificate.
Read the European Commission’s Q&AScoring candidates with AI falls under the AI Act’s high-risk rules from 2 December 2027. A human review, logged evidence per criterion and a versioned rubric are already in place.
Your first three candidates are free. After that you pay per candidate who finishes the test. Anyone who does not start or does not finish costs nothing.