Home · Business · HR & People · Performance & training

PARTLY

As of 13 August 2026, AI can only partly prepare an employee performance review.

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

Can you do it?

5 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 costsNo price for a human performance-review service is provided in the supplied tool information.

If this goes wrong: an inaccurate or biased review damages trust and may contribute to a grievance or an unfair employment decision.

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 employee's objectives, previous review, relevant records and the organisation's review template, and remove information that is not needed for this review.
    2. Gather specific evidence for each objective, including results, dates, examples of work, feedback and any agreed changes to priorities or support.
    3. Ask the employee for their self-assessment and comments, then record their account separately from your own observations.
    4. Paste the prompt and the anonymised objectives, evidence, comments and context into an approved AI tool, then ask it to produce the review draft.
    5. Compare every achievement, gap, date and quotation in the draft against the source records, and delete any claim that cannot be supported.
    6. Ask a colleague or HR contact to check the draft for consistent standards, loaded language, missing context and possible discrimination or process risks.
    7. Send the checked draft through your organisation's review process, and use it as a discussion document rather than treating the AI output as the final decision.

    Prompt

    Prepare a fair, evidence-based employee performance review from the information below.

Employee role: [ROLE]
Review period: [REVIEW PERIOD]
Agreed objectives and measures: [OBJECTIVES]
Evidence of results, with dates or sources where available: [EVIDENCE]
Strengths and positive contributions: [STRENGTHS]
Areas where expectations were not met: [GAPS]
Support, training or resources already provided: [SUPPORT]
Employee self-assessment or comments: [EMPLOYEE COMMENTS]
Relevant context, such as changed priorities or agreed adjustments: [CONTEXT]

Write a professional draft with these headings: overall summary, achievements against each objective, development areas, support and next steps, and discussion points for the review meeting. Separate confirmed evidence from interpretation. Use specific examples and neutral language. Do not invent facts, dates, quotations, causes, diagnoses or legal conclusions. Do not make a recommendation about dismissal, pay, promotion or formal disciplinary action. Flag any missing evidence, conflicting accounts, possible bias, or issue that should be checked with HR before the draft is used. Keep the employee's dignity intact and make the wording suitable for a two-way meeting. End with a short list of factual claims I must verify against the source records.

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

What it gets wrong

  • AI cannot know whether the evidence gives a fair account of the employee's work unless you provide complete records and context.
  • AI cannot resolve conflicting accounts between a manager and an employee.
  • AI cannot decide whether a performance concern reflects capability, unclear expectations, inadequate support or circumstances requiring a different process.
  • AI cannot take responsibility for the employment consequences of the wording or recommendation.
  • AI cannot replace HR or management judgement where the review may lead to formal action, pay decisions or a dispute.

What caps this at PARTLY: legal accountability, judgement under ambiguity and stakes of error.

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
Inputs1
Verification1
Liability1
Effort delta2
Total7 / 10

The methodology and its thresholds are published in full.

FAQ

Can ChatGPT write an employee performance review?
Yes, it can produce a useful first draft from objectives, evidence and employee comments. You still need to check every factual claim and have the wording checked for fairness and process risk before using it.
Is it safe to use AI for employee performance reviews?
It is suitable for drafting when you use an approved tool, minimise personal data and keep a human decision-maker responsible. Do not treat its assessment of performance, fairness or employment consequences as a decision.
What should I give AI to write a performance review?
Give it the review period, agreed objectives, specific evidence, relevant feedback, the employee's own comments and any changed circumstances or support provided. Separate confirmed facts from opinions and tell it not to invent missing details.
Can AI decide if an employee is underperforming?
No. AI can organise evidence against agreed objectives, but it cannot reliably judge disputed context, adequacy of support or whether the process is fair. The manager and employer must make and own that decision, with HR involvement where the consequences are serious.

Nearby answers

All Performance & training rulings

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.