Home · Business · HR & People · Job ads & screening

PARTLY

As of 13 August 2026, AI can only partly recommend which candidates to interview.

This still needs a person who signs their name to it.

Can you do it?

15 minutesto a draft.

30 minutesto something you’d act on.

Cost, all in£0

Skill neededchat-fluent

Who has to check ita colleague

What the alternative costsA purpose-built alternative is Manatal, described as AI recruitment software for candidate scoring, enrichment and pipeline automation.

If this goes wrong: a suitable person is rejected or an unfair pattern enters the shortlist, and the employer has to deal with the employment, reputational and legal consequences.

What to actually do

  1. Hand it to a person

    The route this page recommends

    A person who owns the outcome does this end to end, worth it when the failure is dear.

  2. Use a tool built for this

    Second choice
  3. Do it yourself

    The distant third

    A chat interface, chat-fluent skill, and roughly 30 minutes until you can act on the result.

    How to actually do it

    1. Open the approved job description and selection criteria, then remove vague preferences and separate essential requirements from desirable ones.
    2. Confirm that the employer has a lawful recruitment process and permission to use the application data in an AI system, then remove unnecessary personal information from the documents.
    3. Give each candidate a neutral ID, gather the relevant CVs and application answers, and paste them into the prompt under the matching IDs.
    4. Paste the job description, criteria and candidate material into a chatbot and ask it to produce the evidence-based ranking and human-review flags.
    5. Compare every recommendation with the original essential criteria and the cited application evidence, correcting any invented, missing or irrelevant reasoning.
    6. Ask a second hiring manager or HR professional to check the shortlist for inconsistent treatment, proxy criteria and unexplained exclusions before anyone is invited.
    7. Record the final human decision and the evidence for it, then send interview invitations only through the employer's approved recruitment system.

    Prompt

    You are assisting with a UK recruitment shortlist, not making the hiring decision.
    
    Role: [paste the current job description]
    Essential criteria: [list only criteria genuinely required for the role]
    Desirable criteria: [list these separately]
    Selection process: [describe any approved screening questions or tests]
    Candidates:
    [paste each CV or application under a unique candidate ID]
    
    Rank the candidates for interview using only the role requirements and evidence supplied. Do not use or infer age, sex, gender, race, ethnicity, nationality, religion, disability, pregnancy, marital status, sexual orientation, caring responsibilities, health, name, photograph, home address, school, or any other protected or irrelevant personal characteristic. Do not treat gaps, accents, writing style, employment history, university prestige or location as a proxy for suitability unless the requirement is explicit, job-related and objectively necessary.
    
    Return a table with candidate ID, evidence against each essential criterion, evidence against desirable criteria, missing information, and a recommendation of interview, do not interview, or human review. Give a short reason for every recommendation and quote or clearly identify the application evidence used. Do not invent evidence or assume that missing evidence means the candidate lacks the skill. Flag criteria that are vague, subjective, duplicated, or likely to create unfair bias. Explain any tie and identify cases where a human should review the application before a decision. Finish with a checklist for the hiring manager to verify before inviting anyone to interview. This is not professional advice and must not replace the employer's approved recruitment process or human decision.

    Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.

What it gets wrong

  • AI cannot decide whether a vague requirement is genuinely necessary for the job.
  • AI cannot reliably distinguish a legitimate job-related signal from a proxy for a protected or irrelevant characteristic.
  • AI cannot take responsibility for an unfair shortlist or explain the decision to a candidate, regulator or tribunal.
  • AI cannot supply missing evidence about motivation, communication, reasonable adjustments or context behind a career history.
  • AI cannot replace an accountable hiring manager's judgement about who should be interviewed.

What caps this at PARTLY: legal accountability, judgement under ambiguity and consent and privacy.

How we scored this

Five axes, each scored nought to two by hand: ten means AI carries the task cleanly, and the thresholds that turn a total into YES, PARTLY or NO are published in the methodology. Each axis name links to its definition.

AxisScore (0–2)
Output2
Inputs2
Verification1
Liability0
Effort delta2
Total7 / 10

FAQ

Can AI shortlist candidates for interview?
Partly. It can compare CVs and application answers against a clear, job-related rubric and produce a first-pass shortlist, but a human must check the evidence and own the decision.
Is it legal to use AI to screen job applicants in the UK?
AI screening creates legal and fairness risks, especially where personal data or protected characteristics are involved. This is not professional advice; a serious case needs an employment solicitor or qualified HR professional.
Can AI remove bias from recruitment?
No. It can flag some inconsistent criteria and make its reasoning more visible, but it can reproduce bias in the data, criteria or proxies used. A human still needs to test the process and review borderline cases.
Should I let AI decide who gets an interview?
No. Use it to organise evidence and identify candidates for human review, not to make the final decision. The employer and hiring manager remain accountable for the shortlist.

Nearby answers

Assessed by gpt-5.6-luna (gpt-5.6-luna) on 2026-08-13, second-checked by an independent model. Wrong somewhere? Email [email protected] and it gets re-checked.

The newsletter

AI news, new answers and product picks, straight to your inbox.