AI Pre-Interview Questions by Role: A Catalogue for 8 Roles

What's Included in This AI Pre-Interview Question Catalogue

An AI pre-interview is only as good as its questions. Too many and candidates drop out, too few and nothing can be decided, and vague questions produce answers nobody can compare. This Word file gives you ready-to-use questions for eight common roles, so you do not start from a blank page.

1. Eight roles, five questions each: registered nurse, warehouse operative with forklift licence, customer service, retail sales, hospitality service, inside sales, office and administration, and electrician.
2. Must and nice questions: every question is marked, so the shortlist rests on facts and the conversation afterwards on the rest.
3. A follow-up for each question: what to ask when the first answer is thin, without pushing.
4. A scoring anchor for each question: what a good answer contains. The anchors score content only, never voice, pace or mood.
5. Scoring levels for situation questions: three levels from “no concrete situation” to “situation, own action and outcome”.
6. Transparency text and a legal overview: a notice for the start of the interview, Art. 13 and 22 GDPR, Art. 50 and Art. 5 of the EU AI Act and the anti-discrimination rules, plus a list of questions you do not ask.

Key Features That Save You Time

• Questions that fit the role. A nurse is asked about a deteriorating patient, a warehouse operative about a day with too much at once, an electrician about a fault they traced.
• Objection first. Candidates distrust an AI voice, and decisions about people belong to people. The catalogue is built so that a person always reviews the outcome and the notice says so.
• Facts before stories. Each role starts with the facts that cannot be learned on the job, then one situation question. That keeps the interview to eight to twelve minutes.
• Placeholders instead of invented details. Shift pattern, employer, region and start date are marked in curly braces.
• Fairness built in. The fairness check at the end covers the human review, testing with dialects and accents, and a quarterly look at drop-off and outcomes.

How to Use This AI Pre-Interview Question Catalogue

1. Check first whether an AI interview suits the role at all. The decision tree in our guide to roles for AI interviews helps.
2. Take the five questions for your role and replace the placeholders. Remove any question that is not needed for the job.
3. Keep the must questions to facts and one situation. Move everything else to the nice questions or to the conversation with a person.
4. Use the scoring levels to read answers on their content and flag the ones worth a conversation.
5. Send the invitation, offer a person as an alternative and review every rejection that rests on the interview.

Why This AI Pre-Interview Question Catalogue Works

Structured questions put every candidate on the same footing. That is the core of a fair pre-interview and the reason scoring anchors matter: they tie the assessment to the content of the answer. For the fairness side, read bias and fairness in AI interviews, and for the legal frame what is allowed in AI interviews in the EU.

The legal basics are simple to state. Candidates must know they are talking to an AI system (Art. 50 of the EU AI Act, transparency duties from 2 August 2026). A decision with significant effect, such as a rejection, should not rest solely on automated processing (Art. 22 GDPR). Emotion recognition in the workplace is prohibited, except for medical or safety reasons (Art. 5(1)(f) AI Act). Systems for recruitment and selection are high-risk under Annex III, with deployer obligations from 2 December 2027 as amended by Regulation (EU) 2026/1744.

For a concrete example of how a pre-interview fits into a hiring flow, see from application to interview in ten minutes.

A Closer Look: The Warehouse Operative Questions

To show how the pieces fit, here is the structure of one role. The first question checks a fact that cannot be learned on the job: a valid forklift licence, since when, and for which vehicles. The follow-up asks about the vehicle type, and a good answer names licence, year and type.

The second question asks where the person last worked in a warehouse and what the tasks were. The follow-up asks which goods and what volume. The third is the situation question, a day when a lot came up at once. The anchor looks for the order of priorities, the reasoning and the outcome, not for polish.

The fourth question covers shifts, location and commute, and it is marked must because an early shift that cannot be reached is a real obstacle. The fifth is open, marked nice and not scored: it gives the candidate room and gives you the open points for the conversation afterwards. The whole set takes about ten minutes, and nothing in it depends on how someone sounds.

Who Should Use This

• Recruiting teams that hire in volume and want one question set per role.
• HR business partners who agree questions with hiring managers before the process starts.
• Teams that are setting up their first voice or chat pre-interview and need a solid starting point.
• Data protection officers and works councils who want to see the questions and the notice in advance.

If you want to run pre-interviews in 12 languages without building the flow yourself, look at AI interview and CV screening.

FAQs

How many questions should an AI pre-interview have?

Five is a good number for eight to twelve minutes: facts, one or two situations and a closing question. More questions raise the drop-off, fewer rarely carry a decision.

Can the AI decide who is rejected?

It should not. Art. 22 GDPR limits decisions based solely on automated processing. The pre-interview structures and prepares, a person reviews every rejection.

What should I never ask?

Age, family status, plans for children, pregnancy, religion, health and political views, and nothing that scores voice, expression or mood. The file lists them and gives the legal reason.

Do the questions work for roles that are not in the file?

Yes. Copy the structure: two or three facts that cannot be learned on the job, one situation question with a follow-up, one question about availability and one open closing question.

Is this legal advice?

No. The file is a working aid. Clarify use, data protection and co-determination for your role with your data protection officer.

Which language should the interview use?

The language the candidate is most comfortable in. Voice interviews in Sprad run in 12 languages, so a candidate can answer in their own. Score the content of the answer against the role, not the fluency, unless fluency in a language is a genuine requirement of the job. In that case make it a must question and say so in the invitation.

How should I handle candidates who prefer a person?

Offer it without friction: a reply to the invitation and a call back from a person. Candidates who choose the person must not be scored lower, so the alternative has to be equivalent. Note how many choose it. A high share is a signal that the invitation or the questions need work.