Diagnostic tool

Is your HR AI system high-risk?

Most HR teams deploy AI without knowing that employment and worker management is one of the few uses regulators single out by name. Answer eight questions about one system to get an indicative classification, the obligations likely attached to it, and the ISO/IEC 42001 controls to put in place first.

About 3 minutes · one system at a time · no sign-up required

Question 01 · The system

What does the system do?

Pick the closest match. If it does several things, assess the one with the greatest consequence for an individual.

Question 02 · Consequence

How much does the output shape what happens to a person?

Be honest about practice, not policy. If reviewers accept the output more than about nine times in ten, it is determining the outcome.

Question 03 · Inference

Does it infer things a person never stated?

Emotion, sentiment, engagement, personality, “culture fit,” stress, health, or anything approximating a protected characteristic.

Question 04 · Human oversight

Is there a named person who reviews outputs and can overturn them?

Question 05 · Bias testing

Have outcomes been tested for disparate impact in the last twelve months?

Testing means measuring selection rates across protected groups using your own data — not reading a vendor’s fairness statement.

Question 06 · Documentation

Do you hold documentation of how the system was built and what it should not be used for?

Training data description, intended purpose, known limitations, performance metrics, update history.

Question 07 · Transparency

Are the people affected told the system is used?

Question 08 · Accountability

Does this system appear in a governed inventory?

A named owner, a record of what it does, logged decisions, and a scheduled review date.

Question 09 · Reach

Where are the affected candidates or employees located?

Select all that apply. Obligations follow the person, not your headquarters.