Day 2: Generate Structured Interview Questions

The concept

Interviews should produce evidence. Too often, they produce impressions.

Unstructured interviews can feel natural and useful, but they are vulnerable to inconsistency. One interviewer focuses on technical depth. Another focuses on confidence. Another asks about culture fit without defining what that means. By the time the panel debrief happens, candidates have not been assessed against the same evidence.

AI can help design structured interview questions quickly, but the HR value is not speed alone. The value is consistency, fairness, and better calibration.

Plain English

A structured interview helps the panel compare candidates against the role, not against each other's conversational style.

Start with competencies

Do not ask AI to create interview questions until the competencies are clear.

Competencies should describe what predicts success in this specific role. Avoid generic lists such as "communication, teamwork, problem solving" unless you define them in context.

Better examples:

• influence senior stakeholders without direct authority
• translate ambiguous business problems into workable analysis
• manage high-volume employee relations cases with good judgement
• build trust with managers who are under pressure
• prioritise competing operational requests without losing service quality

The more specific the competency, the better the question.

Translate the role into evidence

A strong interview framework starts with a simple question:

What evidence would convince us this person can do the job at this level?

That question is more useful than "What should we ask?" because it prevents the interview from becoming a list of interesting conversations. The interview should be designed around evidence that predicts performance.

For example, if the competency is "manage employee relations cases with sound judgement," the evidence might include:

• how the candidate gathers facts
• how they separates allegation from interpretation
• how they weighs policy, precedent, and employee impact
• how they documents decisions
• how they escalates risk
• how they communicates with managers under pressure

Those evidence points can become questions, probes, and scoring anchors. Without that step, AI may produce questions that sound professional but do not reveal enough.

Make the interview stage-specific

The same competency can be assessed differently at different stages.

First-round screen

The goal is usually signal, not full diagnosis. Questions should confirm relevant experience, motivation, communication, and basic judgement. Keep them focused.

Hiring manager interview

This stage should test the core work of the role. Questions can go deeper into decision-making, trade-offs, stakeholder situations, and examples of similar complexity.

Panel interview

Panels work best when each interviewer owns a defined area. AI can help create a panel map so three interviewers do not all ask different versions of the same question.

Final stage

Final interviews often test alignment, seniority, operating style, and practical trade-offs. Keep the discussion structured even when the conversation becomes more strategic.

When you fill in the prompt, include the stage. It changes the quality of the output.

Behavioural and situational questions serve different purposes

Behavioural questions ask for past evidence:

Tell me about a time when you had to influence a manager who disagreed with your recommendation.

Situational questions test judgement:

If a hiring manager wanted to reject a candidate for a vague "culture fit" reason, how would you respond?

Both are useful. Behavioural questions show what the candidate has done. Situational questions show how they think through relevant scenarios.

The prompt includes both so the interview does not rely only on polished past stories or only on hypothetical answers.

Probe vague answers

Follow-up probes are essential because candidates often answer at a high level.

Useful probes include:

• What was your specific role?
• What evidence did you use?
• What trade-off did you consider?
• What happened after that?
• What would you do differently now?
• How did you know it worked?

AI can generate probes for each competency, which helps less experienced interviewers go deeper without improvising poorly.

Build the rubric before interviews begin

Questions are only half the system. Scoring matters.

Without a rubric, interviewers may rate confidence, similarity, polish, or personal preference. A rubric defines what weak, acceptable, and strong evidence looks like.

For each competency, define:

• 1 out of 5: little relevant evidence or poor judgement
• 3 out of 5: adequate evidence with some gaps
• 5 out of 5: strong, specific evidence at the required level

This makes debriefs more disciplined. The panel can discuss evidence rather than vibes.

What good interview notes look like

Structured interviews only work if the notes support the decision. Notes should capture evidence, not personality labels.

Weak note:

Great energy. Seems like a strong fit.

Better note:

Described reducing time-to-fill for specialist roles by changing the intake process, clarifying must-have criteria, and introducing weekly hiring manager updates. Provided metrics and explained one trade-off around candidate quality.

The better note can be discussed. The weak note cannot.

Encourage interviewers to write notes in three buckets:

• Situation: what context did the candidate describe?
• Action: what did the candidate personally do?
• Outcome: what changed, and how do they know?

This keeps the conversation anchored in evidence.

Watch for bias in scoring

AI can flag common bias risks, but the panel must manage them.

Bias can enter when interviewers:

• overvalue confidence
• confuse polish with competence
• reward familiar career paths
• penalise different communication styles
• use "culture fit" without definition
• compare candidates to each other instead of the rubric
• let one strong answer influence unrelated scores

Mitigation is practical: use the same questions, take evidence-based notes, score independently before discussion, and require examples for claims.

Calibrate before the first candidate

A hiring panel should not discover its standards during the debrief. Calibration should happen before interviews begin.

Share the AI-generated question bank and rubric with the panel. Then ask:

• What would a strong answer sound like?
• What would be acceptable but not exceptional?
• What answer would create concern?
• What evidence is required for a high score?
• Which competencies are most important if trade-offs appear?

This does not remove judgement. It improves judgement by giving it a shared frame.

For senior roles, calibration is especially important because candidates may have different styles. One candidate may be concise and direct. Another may be more reflective. Another may have worked in a different industry. The panel must decide whether the evidence meets the role requirements, not whether the candidate tells the story in the most familiar way.

Candidate questions are part of the experience

The prompt asks AI to prepare likely candidate questions and suggested responses. This is not a minor detail.

Candidates often ask about team culture, success expectations, flexibility, growth, decision timelines, and why the role is open. If interviewers give inconsistent answers, the candidate experience suffers and trust declines.

Prepare clear responses to common questions. If the answer is not yet known, say so honestly and explain when it will be clarified. A polished answer is less important than a truthful one.

Today's practice

Choose one role. Define three to five competencies. Run the prompt. Then ask the hiring manager:

1. Do these competencies predict success here?
2. Which questions would reveal useful evidence?
3. Which questions feel generic or irrelevant?
4. What would a strong answer look like at this level?
5. Where might the panel score inconsistently?

By the end, you should have an interview framework that supports better hiring decisions and a more defensible process.