The Return on People

Templates · One page · Not legal advice

4 questions for your HR tech vendor.

Ask these before you buy or switch on AI for hiring, promotion or performance. Send them by email, so the answers are in writing. In the EU, AI used for hiring is high risk under the AI Act, and the main hiring obligations apply from 2 December 2027. That makes these questions worth asking now.

1. What data trained it?

Ask: whose data, from which countries and which years, and for which kinds of jobs.

Why it matters: a tool trained on past hiring decisions can learn the patterns of those decisions, including the unfair ones.

2. Was it tested for bias, and by whom?

Ask: for which groups, with what result, and was the test done by the vendor or by someone independent. Ask to see the report.

Why it matters: "we tested it" means little without the groups, the numbers and who did the testing.

3. Can a human override it?

Ask: who makes the final decision, how a recruiter changes the tool's result, and whether that is logged.

Why it matters: in a large field experiment on AI job interviews, a human recruiter made every hiring decision. Jabarian & Henkel, 2025. Keep that design. More on AI in hiring

4. Can a candidate ask why?

Ask: what a rejected candidate can be told, and how they can ask for a person to look again.

Why it matters: people judge a process as fair partly by whether they get an explanation and a voice. That is a long standing finding in work psychology on procedural justice.

Before you send it

Tell candidates up front when AI is used, and offer a human option. This template is a starting point for a conversation, not legal advice. For your legal duties, ask your legal or data protection team.

Free guide, 6 pages

AI at work, without the burnout.

Seven rules from the newest research. You also get new studies by email, never more than one a week.

No spam. Unsubscribe with one click, any time. Privacy · Read past letters