Course › Module 7 · Structured and AI interviews

Lab 6: run one structured interview loop

Module 7, Lesson 7  ·  3 min read ·  Updated 21 September 2026

Module 7 · Lesson 7

Design and run one complete interview process, start to finish, and produce scorecards someone else could check.

What to do

  1. Take the scorecard from Module 2 and the shortlist from Module 5.
  2. Design the process: which skills get checked at which stage, by whom, and how long each stage takes.
  3. Build the questions — two questions and three follow-ups for each skill you test at interview.
  4. Write what each score means for each skill, if you have not already.
  5. Run one interview. A colleague playing the part of one of the practice candidates works fine.
  6. Fill in the scorecard from the evidence, using the method in Lesson 5.
  7. Hold a discussion — even with one other person — using the structure from Lesson 6.

What to hand in

The process design, the questions with follow-ups, one completed scorecard with a quote for each skill, and a short note on which skill was hardest to find evidence for.

How to know you have done it right

CheckGood looks like
Everything is coveredEvery skill on the scorecard gets checked somewhere
Nothing is doubled upNo skill checked at three stages while another has none
Questions trace backEvery question matches a skill
EvidenceEvery score quotes something the candidate said
Honest gaps"No evidence" written where nothing came up — not a middle score
Score descriptions usedYou can say why a 3 and not a 2, pointing at what you wrote

Check yourself before Module 8

  • Which skill was hardest to find evidence for? That usually means it belongs at a different stage — or cannot be judged in an interview at all.
  • Did any follow-up question change your view of an answer? If none did, your follow-ups are too gentle.
  • Could a second interviewer, given your questions and score descriptions, arrive at similar scores?
  • Did you write anything down that describes the interview rather than the job? Those are the lines to cut.

How long to spend

Ninety minutes including the interview. The process design and the score descriptions are reusable for every future hire into this kind of job, which is where the return on this exercise actually sits.

Common questions

Are structured interviews actually better?

The research consistently finds structured interviews predict job performance substantially better than unstructured ones. The specific effect sizes from older meta-analyses were overstated and have been revised, but the direction has survived every serious re-analysis. What unstructured interviews mostly measure is rapport, early impression, and interview skill.

Should you use AI to score video interviews?

Scoring what a candidate said against your criteria is defensible. Scoring appearance, facial expression, tone or speech rhythm is not — the inference is weakly supported and it disadvantages people by disability, neurotype and first language. It is also the single most regulated feature in the stack, and at least one major vendor withdrew facial analysis after external review.

How do you stop an interview debrief becoming groupthink?

Lock every interviewer's scores and evidence before the meeting opens — not 'please avoid discussing it beforehand', which nobody manages. Then run competency by competency rather than person by person, spend the time on divergence, have the least senior speak first, and require evidence before opinion.

Check yourself

Four questions. Nothing is recorded anywhere but your own browser — this is for you, not for a score.

  1. 1What does an unstructured interview mostly measure?

  2. 2Which probe most reliably separates someone who did the work from someone who was nearby?

  3. 3A vendor says their system scores candidates on facial expression and tone. What is the issue?

  4. 4Why must interview scores be locked before the debrief?