Reviewed and updated: 21 July 2026.
Recorded interview notes can improve recruitment evidence, but only when the scorecard is designed before the interview and the final assessment remains job-related, consistent and human-led. A transcript should support the panel’s memory, not become an automated hiring decision.
Quick verdict: define role criteria and scoring anchors in advance, ask every candidate comparable questions, map evidence to timestamps, score independently and retain meaningful human review throughout.
Commercial disclosure: NERALVO sells the Halo AI voice recorder and uses its DOWAY transcription workflow as an example.
What a recruitment scorecard should do
A scorecard should help trained interviewers compare candidates against the requirements of the role rather than memory, confidence, similarity or general impressions. It should show:
- The criterion being assessed
- The question or exercise used
- The evidence supplied by the candidate
- The scoring anchor applied
- The interviewer’s rationale
- Any follow-up evidence still required
Design the criteria before interviews begin
Start with the job analysis and approved role profile. Choose a limited set of criteria that genuinely predict successful performance, such as:
- Relevant technical knowledge
- Problem solving
- Stakeholder communication
- Planning and prioritisation
- Accuracy and risk awareness
- Leadership or collaboration where required
- Role-specific values or behaviours
Avoid vague categories such as “culture fit,” “executive presence” or “seems confident” unless they are replaced by observable, job-related behaviour.
Create scoring anchors
| Score | Evidence anchor |
|---|---|
| 1 | No relevant example, or answer does not address the criterion |
| 2 | Partial example with limited personal contribution or outcome |
| 3 | Clear relevant example showing appropriate action and outcome |
| 4 | Strong evidence, sound judgement, measurable result and learning |
| 5 | Exceptional evidence exceeding the role requirement in a directly relevant way |
Adapt anchors to each criterion. Interviewers should know what good evidence looks like before meeting candidates.
Use structured questions
Ask all candidates the same core questions in the same order where practical, with consistent follow-up prompts. Useful evidence prompts include:
- What was the situation?
- What responsibility did you personally hold?
- What action did you take?
- Why did you choose that approach?
- What was the result?
- What would you change next time?
Consistent structure improves comparability without preventing reasonable clarification.
Explain recording and data use clearly
Before recording, use the organisation’s approved recruitment process to explain:
- That the interview or interviewer notes will be recorded
- Why recording is being used
- How the transcript will support assessment
- Who can access it
- How long it will be retained
- Whether any automated tools are involved
- How the candidate can raise questions or exercise applicable rights
- What alternative process is available where appropriate
The organisation should identify and document an appropriate lawful basis, provide the required privacy information and avoid collecting more candidate data than the recruitment purpose needs.
Capture evidence, not personality labels
| Criterion | Candidate evidence | Timestamp | Provisional score |
|---|---|---|---|
| Prioritisation | Replanned three competing deadlines, agreed trade-offs with stakeholders and delivered the highest-risk item first | 18:42 | 4 — subject to panel review |
Do not replace evidence with labels such as “natural leader,” “not polished” or “does not sound senior.”
Correct the transcript before scoring
Interview transcripts can misidentify:
- Candidate and interviewer speakers
- Names
- Technical terms
- Negative words
- Numbers and dates
- Accents or speech patterns
- Answers spoken over another person
Replay the relevant audio before relying on disputed wording. The scorecard should never depend on an unchecked transcript fragment.
Do not infer traits from voice or delivery
A recording may capture accent, pace, pauses, disability-related speech differences, nervousness or connection quality. These are not reliable proxies for competence. Do not ask AI to infer honesty, emotion, personality, ethnicity, health, confidence or likely performance from the voice.
Separate relevant evidence from protected or personal information
Candidates may disclose age, health, family circumstances, religion or other personal information while answering. Exclude irrelevant details from the decision record and assess only the approved role criteria. Escalate any necessary adjustment or legal issue through the organisation’s trained HR process.
Support reasonable adjustments
A structured interview should allow agreed adjustments without lowering the role standard. Examples may include extra processing time, alternative communication formats, accessible technology, rest breaks or a different assessment method. Record the evidence fairly rather than penalising the adjustment.
Score independently before panel discussion
- Each interviewer reviews the relevant evidence.
- Each interviewer completes a provisional score and rationale independently.
- The panel compares scores criterion by criterion.
- Differences are resolved using evidence, not seniority or group pressure.
- The final rationale is documented.
Independent scoring reduces the risk that the first or most senior interviewer defines the group’s view.
Keep meaningful human review
AI may draft transcripts, group answers or suggest where evidence appears. It should not make the final hiring decision, rank candidates without transparent controls or override the trained panel. A reviewer should have genuine authority, sufficient information and time to challenge the output.
Check current official guidance
The Information Commissioner’s Office recruitment and selection guidance covers candidate information and automated decision-making considerations. The ICO’s Recruitment Rewired work also highlights regulatory scrutiny of AI recruitment tools. The Equality and Human Rights Commission has warned that poorly implemented AI can perpetuate discrimination. Review current guidance and obtain specialist advice for the organisation’s process.
Monitor quality and fairness
Recruitment teams should review:
- Transcript error rates
- Missing or mislabelled speakers
- Score consistency between interviewers
- Pass rates across relevant groups where lawful and appropriate
- Candidate complaints or adjustment failures
- Whether criteria still match the role
- Whether the tool changes who progresses
Unexpected patterns require investigation, not automatic acceptance.
Control access and retention
Store recruitment recordings, transcripts and scorecards only in approved systems. Limit access to the recruitment team, record who reviewed them and delete source material according to the organisation’s documented retention schedule unless a justified need requires longer storage.
Using NERALVO Halo
NERALVO Halo provides NOTE mode, CALL mode for supported and permitted calls, 64GB local storage, up to 35 hours of recording and Bluetooth connection to DOWAY. DOWAY can generate transcripts, summaries, templates, translations, mind maps and exports, with one year of DOWAY Max included.
Use Halo only within the employer’s approved recruitment process. Verify the transcript, transfer relevant evidence into the authorised scorecard and keep the hiring judgement human-led.
Recruitment scorecard checklist
- Role criteria defined before interviews
- Scoring anchors describe observable evidence
- Core questions are consistent
- Recording and data use are explained
- Lawful basis and privacy information are documented
- Transcript errors are corrected
- Evidence is mapped to timestamps
- Voice, accent and emotion are not treated as competence
- Irrelevant personal information is excluded
- Reasonable adjustments are supported
- Interviewers score independently
- Meaningful human review is retained
- Access and retention are controlled
- Fairness and quality are monitored
Frequently asked questions
Can AI score candidates automatically?
Automated scoring can introduce legal, fairness and accuracy risks. Use structured human assessment and obtain specialist review before introducing any automated decision-making.
Should the complete transcript go to the panel?
Only where necessary and authorised. A verified evidence record linked to relevant timestamps is often more usable and proportionate.
Can interviewers score communication style?
Only where a clearly defined communication capability is genuinely required for the role and assessed using job-related evidence rather than accent, personality or presentation preference.
What if a candidate declines recording?
Follow the organisation’s approved alternative process and explain how comparable notes will be created.
Make evidence comparable without automating judgement
A strong scorecard combines consistent questions, verified evidence, transparent data handling and trained human judgement. Recording improves recall only when the recruitment process remains fair and accountable.
Ready to capture meetings properly?
View the NERALVO Halo AI voice recorder with 64GB local storage, meeting capture, compatible phone-call recording workflows and one year of DOWAY Max included.
View NERALVO Halo