ats for startups

AI interview notes: keep judgement with people

AI interview notes work best when they capture what was said and leave the hiring decision with trained interviewers.

A recruiter with open hands as spoken comments in speech bubbles gather into a tidy ticked notes card

Where AI helps without deciding

AI interview notes earn their place when they capture what was said and leave the decision to people. Recording, transcribing and organising an interview is admin. Assessing the answers, weighing them against the role and explaining the outcome is judgement. Hiring teams that keep the two apart get a better record of each interview and a decision they can explain.

That split is a practical rule of thumb and not a legal test. The regulator’s concern is decisions. The UK Information Commissioner’s Office (ICO) describes solely automated recruitment decisions as those made without meaningful human involvement, and treats shortlisting, recommending for interview and rejecting candidates as decisions with significant effects. A tool that produces a transcript does not make any of those decisions. A tool that scores or recommends candidates does, so it needs a closer look.

What to check in any AI feature

AI note-taking products vary in what they do with a recording. Before switching a feature on, ask what it infers and from what. The ICO suggests asking providers about training data, bias testing and fairness testing, and the questions below turn that into a short supplier checklist.

  • Does the feature output a transcript and summary of what was said, or an assessment of the candidate?

  • Does it analyse personality, behaviour or other traits? The ICO describes profiling as analysis of exactly those kinds of characteristics, and it brings extra rules.

  • Has the supplier tested for accuracy and bias? The ICO says to regularly review your AI system and have measures in place to check for bias or discrimination.

  • Can candidates who need adjustments take part fairly? The ICO says you must make reasonable adjustments for people who have a disability, and Acas says employers must ask anyone coming to an interview whether they need reasonable adjustments.

Being able to explain a decision

If you cannot explain a decision to a candidate, you will struggle to explain it to anyone else. Candidates can make a subject access request that covers the outcomes and results relating to them at every stage, so interview records need to be accurate and fit to be read.

A person raises a finger beside a ticked notes card and two options, ready to make a decision.

Acas advises asking each applicant the same questions where possible, having more than one interviewer, basing decisions on the job description and person specification, getting permission before recording, and storing notes and recordings securely. CIPD adds that interviewers should score each response independently before group discussion, because discussing first can sway people, including towards the most senior person present. A clean record of what was asked and answered supports all of this, while the scoring and the decision stay with the interviewers.

Practical steps

These steps are mostly conversations with your supplier and your interviewers, and each one can be done before the next hiring round.

  • Get permission before recording, and tell candidates what happens to the recording and notes, including how long they are kept. The ICO says not to keep recruitment information just in case.

  • Capture and organise with AI, and keep scoring with interviewers against criteria agreed before the interviews. This division of work is our suggestion rather than a requirement.

  • Complete a data protection impact assessment (DPIA) before using any AI that helps make recruitment decisions. The ICO says one is mandatory for solely or partly automated decision-making and profiling in recruitment.

  • Spot check transcripts and summaries against what was actually said, and correct errors before they reach a debrief.

Dividing the work between AI and interviewers. AI handles: capture and transcribe what was said; organise questions and answers for the debrief. Interviewers keep: assessment against shared criteria; the decision and the explanation for it. A practical rule of thumb. General information only

How Screenloop fits

Screenloop’s Interview Intelligence records and transcribes interviews and pulls out the questions asked and the answers given, so the debrief starts from what was said. It handles capture and organisation. Scoring and decisions stay with the interviewers.

Common questions

Do we need candidate permission to record an interview?

Acas says employers must get a person’s permission before recording an interview, and should store notes and recordings securely and confidentially under data protection rules.

Is an AI note-taker the same as automated decision-making?

The ICO’s guidance concerns decisions, and it defines solely automated decision-making as decisions made without meaningful human involvement. A tool that captures a transcript does not decide anything. Any feature that scores, ranks or recommends candidates falls within the ICO’s concerns, so check each feature separately.

This article is general information, not legal advice. ICO guidance on automated decision-making in recruitment is under review, so check the current version before you act.

Contributors

Screenloop

Editorial Team

Contributors

Screenloop

Editorial Team

Subscribe to our newsletter

Where AI helps without deciding

AI interview notes earn their place when they capture what was said and leave the decision to people. Recording, transcribing and organising an interview is admin. Assessing the answers, weighing them against the role and explaining the outcome is judgement. Hiring teams that keep the two apart get a better record of each interview and a decision they can explain.

That split is a practical rule of thumb and not a legal test. The regulator’s concern is decisions. The UK Information Commissioner’s Office (ICO) describes solely automated recruitment decisions as those made without meaningful human involvement, and treats shortlisting, recommending for interview and rejecting candidates as decisions with significant effects. A tool that produces a transcript does not make any of those decisions. A tool that scores or recommends candidates does, so it needs a closer look.

What to check in any AI feature

AI note-taking products vary in what they do with a recording. Before switching a feature on, ask what it infers and from what. The ICO suggests asking providers about training data, bias testing and fairness testing, and the questions below turn that into a short supplier checklist.

  • Does the feature output a transcript and summary of what was said, or an assessment of the candidate?

  • Does it analyse personality, behaviour or other traits? The ICO describes profiling as analysis of exactly those kinds of characteristics, and it brings extra rules.

  • Has the supplier tested for accuracy and bias? The ICO says to regularly review your AI system and have measures in place to check for bias or discrimination.

  • Can candidates who need adjustments take part fairly? The ICO says you must make reasonable adjustments for people who have a disability, and Acas says employers must ask anyone coming to an interview whether they need reasonable adjustments.

Being able to explain a decision

If you cannot explain a decision to a candidate, you will struggle to explain it to anyone else. Candidates can make a subject access request that covers the outcomes and results relating to them at every stage, so interview records need to be accurate and fit to be read.

A person raises a finger beside a ticked notes card and two options, ready to make a decision.

Acas advises asking each applicant the same questions where possible, having more than one interviewer, basing decisions on the job description and person specification, getting permission before recording, and storing notes and recordings securely. CIPD adds that interviewers should score each response independently before group discussion, because discussing first can sway people, including towards the most senior person present. A clean record of what was asked and answered supports all of this, while the scoring and the decision stay with the interviewers.

Practical steps

These steps are mostly conversations with your supplier and your interviewers, and each one can be done before the next hiring round.

  • Get permission before recording, and tell candidates what happens to the recording and notes, including how long they are kept. The ICO says not to keep recruitment information just in case.

  • Capture and organise with AI, and keep scoring with interviewers against criteria agreed before the interviews. This division of work is our suggestion rather than a requirement.

  • Complete a data protection impact assessment (DPIA) before using any AI that helps make recruitment decisions. The ICO says one is mandatory for solely or partly automated decision-making and profiling in recruitment.

  • Spot check transcripts and summaries against what was actually said, and correct errors before they reach a debrief.

Dividing the work between AI and interviewers. AI handles: capture and transcribe what was said; organise questions and answers for the debrief. Interviewers keep: assessment against shared criteria; the decision and the explanation for it. A practical rule of thumb. General information only

How Screenloop fits

Screenloop’s Interview Intelligence records and transcribes interviews and pulls out the questions asked and the answers given, so the debrief starts from what was said. It handles capture and organisation. Scoring and decisions stay with the interviewers.

Common questions

Do we need candidate permission to record an interview?

Acas says employers must get a person’s permission before recording an interview, and should store notes and recordings securely and confidentially under data protection rules.

Is an AI note-taker the same as automated decision-making?

The ICO’s guidance concerns decisions, and it defines solely automated decision-making as decisions made without meaningful human involvement. A tool that captures a transcript does not decide anything. Any feature that scores, ranks or recommends candidates falls within the ICO’s concerns, so check each feature separately.

This article is general information, not legal advice. ICO guidance on automated decision-making in recruitment is under review, so check the current version before you act.

Where AI helps without deciding

AI interview notes earn their place when they capture what was said and leave the decision to people. Recording, transcribing and organising an interview is admin. Assessing the answers, weighing them against the role and explaining the outcome is judgement. Hiring teams that keep the two apart get a better record of each interview and a decision they can explain.

That split is a practical rule of thumb and not a legal test. The regulator’s concern is decisions. The UK Information Commissioner’s Office (ICO) describes solely automated recruitment decisions as those made without meaningful human involvement, and treats shortlisting, recommending for interview and rejecting candidates as decisions with significant effects. A tool that produces a transcript does not make any of those decisions. A tool that scores or recommends candidates does, so it needs a closer look.

What to check in any AI feature

AI note-taking products vary in what they do with a recording. Before switching a feature on, ask what it infers and from what. The ICO suggests asking providers about training data, bias testing and fairness testing, and the questions below turn that into a short supplier checklist.

  • Does the feature output a transcript and summary of what was said, or an assessment of the candidate?

  • Does it analyse personality, behaviour or other traits? The ICO describes profiling as analysis of exactly those kinds of characteristics, and it brings extra rules.

  • Has the supplier tested for accuracy and bias? The ICO says to regularly review your AI system and have measures in place to check for bias or discrimination.

  • Can candidates who need adjustments take part fairly? The ICO says you must make reasonable adjustments for people who have a disability, and Acas says employers must ask anyone coming to an interview whether they need reasonable adjustments.

Being able to explain a decision

If you cannot explain a decision to a candidate, you will struggle to explain it to anyone else. Candidates can make a subject access request that covers the outcomes and results relating to them at every stage, so interview records need to be accurate and fit to be read.

A person raises a finger beside a ticked notes card and two options, ready to make a decision.

Acas advises asking each applicant the same questions where possible, having more than one interviewer, basing decisions on the job description and person specification, getting permission before recording, and storing notes and recordings securely. CIPD adds that interviewers should score each response independently before group discussion, because discussing first can sway people, including towards the most senior person present. A clean record of what was asked and answered supports all of this, while the scoring and the decision stay with the interviewers.

Practical steps

These steps are mostly conversations with your supplier and your interviewers, and each one can be done before the next hiring round.

  • Get permission before recording, and tell candidates what happens to the recording and notes, including how long they are kept. The ICO says not to keep recruitment information just in case.

  • Capture and organise with AI, and keep scoring with interviewers against criteria agreed before the interviews. This division of work is our suggestion rather than a requirement.

  • Complete a data protection impact assessment (DPIA) before using any AI that helps make recruitment decisions. The ICO says one is mandatory for solely or partly automated decision-making and profiling in recruitment.

  • Spot check transcripts and summaries against what was actually said, and correct errors before they reach a debrief.

Dividing the work between AI and interviewers. AI handles: capture and transcribe what was said; organise questions and answers for the debrief. Interviewers keep: assessment against shared criteria; the decision and the explanation for it. A practical rule of thumb. General information only

How Screenloop fits

Screenloop’s Interview Intelligence records and transcribes interviews and pulls out the questions asked and the answers given, so the debrief starts from what was said. It handles capture and organisation. Scoring and decisions stay with the interviewers.

Common questions

Do we need candidate permission to record an interview?

Acas says employers must get a person’s permission before recording an interview, and should store notes and recordings securely and confidentially under data protection rules.

Is an AI note-taker the same as automated decision-making?

The ICO’s guidance concerns decisions, and it defines solely automated decision-making as decisions made without meaningful human involvement. A tool that captures a transcript does not decide anything. Any feature that scores, ranks or recommends candidates falls within the ICO’s concerns, so check each feature separately.

This article is general information, not legal advice. ICO guidance on automated decision-making in recruitment is under review, so check the current version before you act.

Table of Contents

Contributors

Screenloop

Editorial Team

Subscribe to our newsletter