What is an AI interview?
An AI interview is a structured interview conducted live by an agent, in voice or video, at a time that suits the candidate. The agent asks questions drawn from a defined plan and follows up within set limits, while the whole conversation is recorded. The responses are then evaluated against the same rubric applied to every other candidate, with the scores and the supporting evidence passed to people who make the decision.
Also called: Automated interview, Agent-led interview
Why consistency is the main argument for an agent-led interview
Human screening interviews vary with the interviewer and with the order in which candidates are seen. Two applicants for the same place can face different questions and different standards without anyone intending it. An agent asks comparable questions every time and applies the same rubric to every answer. That consistency only counts when the scoring is reviewed for bias and people retain the final decision.
- Every candidate faces a comparable set of questions and follow-ups.
- Interviews can be scheduled at any hour without panel availability.
- Recorded responses allow the scoring to be audited after the event.
- Final decisions remain with people who can weigh context the rubric misses.
How an AI interview works
- The candidate is told how the interview works and how it will be assessed.
- The agent asks questions from a defined plan aligned to the rubric.
- Follow-up questions are generated within limits set in advance.
- The full interview is recorded and transcribed.
- Responses are scored against the rubric with evidence linked to each score.
- Scores and evidence pass to human reviewers who make the decision.
How an AI interview is evaluated
AI interviews are evaluated on agreement with expert judgement and on fairness across candidate groups. Measure the following on a reviewed sample:
- agreement between interview scores and expert panel scores
- score differences between candidate groups with similar qualifications
- consistency of questions asked across candidates
- share of scores supported by linked evidence
- candidate feedback on clarity and fairness of the process
Limitations, risks and common failure modes
- Accent and fluency influence scores that should measure substance.
- Follow-up questions wander and candidates receive different interviews.
- Candidates are not told an agent is conducting the interview.
- Reviewers accept the scores without examining the linked evidence.