Free assessment

Is your human-in-the-loop actually checking?

Score your team's AI oversight in under 5 minutes. Get an instant result, your weakest point, and what to fix first.

12 questions
Instant score
No sign-up

Sample result

74%

Partial oversight

Ready when you are

Answer for the team that reviews your most consequential AI-assisted decisions, such as screening, claims or underwriting. Your answers stay in your browser unless you choose to save your score.

Why the human check fails

68%

Professionals with no AI training in the past year

LSE survey, Lordan 2026

40%

AI-assisted tasks workers said got no critical scrutiny

Microsoft Research and Carnegie Mellon

1983

The year the ironies of automation were first described

Bainbridge, Automatica

Governments have not settled who regulates AI. The EU has delayed its high-risk rules to the end of 2027 and the UK regulates at the point of use. In practice, the control is the person who signs off the AI's output. Automation bias is strongest when a system is usually right, and generative AI is usually right and always fluent. Under UK GDPR, a decision without meaningful human involvement is treated as solely automated.

The four conditions we score

Handoff clarity

The person approving AI decisions knows why a decision reached them and whether it is riskier than the last ten.

Verification habit

Checking means testing the output against something, not reading it for plausibility.

Human capability

Your team is trained on how AI fails and keeps the underlying skill sharp.

Working conditions

People have the time to check, and catching an error is rewarded rather than seen as slowing things down.

Sources and method

This is a self-assessment, not a validated psychometric instrument. Scores indicate where to look, not a compliance finding.