Pre-hire AI assessment · Fig. 01
Now you can measure it.
Each candidate takes a short, role-specific assessment and one real task with a live AI assistant. It measures what they know and how well they work with it, and gives you a clear level from 1 to 5 with the evidence behind it. Run it on your shortlist before you hire, or on the team you already have.
Grounded in OECD, UNESCO and Anthropic research
ChatGPT
Claude
Gemini
CopilotGrounded in
OECD
UNESCO
ECAnthropicThe report
A. Vandenberghe
Assessment report
Builder
Level 4 of 5
Knowledge 40% · Practical 60%
“Where does the 23% figure come from? It is not in anything I gave you, so take it out.”
Caught an unsupported number in the AI's draft and removed it before submitting.
The assessment, live
A basic baseline, the knowledge quiz, the practical with a real AI assistant, and the level you receive — it plays on its own below.
Which tools do you actually use?
ChatGPTUses it
ClaudeKnows it
GeminiKnows it
Microsoft CopilotNever heardHow it works
I
Drop in the job post. AI builds a role-specific assessment in minutes.
II
Eighteen timed questions, then a live practical with a real AI chat.
III
A level from 1 to 5, rubric scores with verbatim quotes behind each.
The scale
Every candidate lands on the same five-level scale, whatever the role. The same yardstick across your whole pipeline.
I
Knows what AI can and cannot do: limits, hallucination, basic ethics.
II
Prompts effectively, evaluates output, picks the right tool for the job.
III
Delegates multi-step work, manages context, verifies the result.
IV
Turns prompts into reusable assets, builds mini-tools and agent workflows.
V
Designs agent systems, weighs risk, governs how a team uses AI.
The foundations · Fig. 02
Three published frameworks define what the five levels measure, and the test itself is built to assessment-science standards. All of it is free to read, so every score traces back to work anyone can check.
OECD · European Commission
Twenty-two AI-literacy competences in four domains: engaging, creating, managing and designing with AI. The domains map almost one to one onto the five-level ladder.
UNESCO
The global reference for what AI competence means across roles and levels. Its progression from understanding to applying to creating runs through the level ladder.
Anthropic
Four competencies, Delegation, Description, Discernment and Diligence, each defined as observable behavior. That is what makes a live chat transcript scorable.
Written to standard
Every question is built to established assessment-design standards and passes an independent review before any candidate sees it, so scores hold up to scrutiny.
More than self-report
The entry-level baseline rests on validated AI-literacy scales and never takes a self-rating at face value: it is paired with objective checks.
Scored on real work
The practical is a situational-judgment task graded on observable behaviour, with a verbatim quote behind every criterion. No box-ticking.
Kept currentThe question bank and scoring rubric are reviewed against the latest AI research and EU AI Act fairness practice, so what you measure reflects how AI is worked with today.
EU AI Act · Article 4
Organisations that provide or use AI systems must ensure their staff have a sufficient level of AI literacy. The obligation has applied since 2 February 2025, and enforcement is phasing in across the EU. No certification standard is mandated.
Measuring AI skill, in your hires and your team, is the cleanest way to comply.
Read the regulationHiring and workforce
The same assessment works two ways. Send it to a shortlist before you interview, or run it across the people already on your payroll.
For hiring
Send a shortlist one link. Get back a level, a rubric, and the evidence behind it, so you walk into every call already knowing who can do the work.
For your team
Baseline where your staff actually stand, find the gaps worth training, and keep a record you can show. Staff AI literacy has been an EU requirement since February 2025.
Compliance · Fig. 03
Candidate assessment is regulated work. GDPR and the EU AI Act shape how this product stores data, scores work and keeps people in charge of the decision.
EU hosting
Candidate data lives on EU infrastructure. The database runs in an EU Supabase region and our functions run in Frankfurt.
No training
Assessments run on Anthropic’s API under a data processing agreement with standard contractual clauses. Candidate data is never used to train models.
Human review
The AI score is decision support. Assessors can override every criterion with a logged reason, designed around GDPR Article 22.
Minimization
The practical collects only the work itself. Prompts avoid personal probing, so the transcript stays about the task.
Erasure
Candidate data is deleted on request: transcripts, scores and deliverables included.
AI Act
The Article 4 AI-literacy duty for employers has applied since 2 February 2025. The product is built to high-risk-system standards, with versioned rubrics, logging and human oversight.
Pricing
From €23.80 per candidate. Credits only burn when someone completes.
Starter
I10credits
€29.00 / candidate
€290 per pack, one credit per completed assessment.
Recommended
Team
II50credits
€23.80 / candidate
€1,190 per pack, one credit per completed assessment.
Enterprise
IIICustom
For volume hiring and rollouts across teams.
Invites that go unanswered cost nothing.
Candidate data stays in the EU. See how we handle it ↑
