Home · Business · HR & People · Performance & training

PARTLY

As of 13 August 2026, AI can only partly write constructive employee feedback.

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 costsThe supplied data gives no reliable price for a human HR adviser or feedback-writing service.

If this goes wrong: the feedback misstates events, appears biased or escalates a performance issue without fair process.

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 relevant objectives, role expectations, notes and work records, and gather only the facts directly related to the feedback.
    2. Write down the observable behaviour or result, its effect, the expected standard and the practical improvement needed, separating evidence from your interpretation.
    3. Remove unnecessary personal information and do not include protected characteristics, medical information or speculation about motives.
    4. Paste the supplied prompt into ChatGPT, Claude or Gemini and replace each bracketed slot with the facts you gathered.
    5. Compare the draft message line by line with your records, removing any claim, example or implication that is not supported by evidence.
    6. Ask a trusted colleague or HR contact to check whether the wording is specific, proportionate and free from obvious bias, then amend the draft.
    7. Deliver the feedback through your normal management process and record the agreed next step and the employee's response.

    Prompt

    Write constructive employee feedback for a UK workplace using the information below. Use only the facts I provide and do not invent examples, motives, dates or outcomes. Keep the tone direct, respectful and specific. Describe the observed behaviour or result, its impact, the expected standard, and a practical next step. Separate facts from interpretation. Do not diagnose the employee, comment on protected characteristics, make legal conclusions, or recommend disciplinary action. If the information is too thin or ambiguous, list the questions I need to answer before sending the feedback. Produce: 1) a concise feedback message addressed to the employee, 2) three factual points I should check against my records, and 3) one question I can ask the employee to hear their perspective. Context: [employee's role and relevant responsibilities]. Situation: [what happened, using observable facts]. Expected standard: [the agreed standard or objective]. Impact: [specific effect on work, customers, colleagues or deadlines]. Previous discussion or support: [relevant factual context]. Desired next step: [what should happen next].

    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 your account of an incident is complete or whether relevant context has been left out.
  • AI cannot decide whether informal feedback, performance support or a formal process is appropriate.
  • AI cannot reliably recognise every form of bias or assess the employment-law risk in the surrounding situation.
  • AI cannot replace the manager's responsibility to listen to the employee and respond to their explanation.
  • AI cannot make the feedback fair by itself when the underlying evidence is weak or inconsistent.

What caps this at PARTLY: judgement under ambiguity, legal accountability and context depth.

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 employee feedback?
Yes, it can produce a clear first draft from specific observations, expected standards and agreed next steps. You must check every factual claim and own the judgement about fairness, process and consequences.
Is it OK to use AI to write performance feedback?
It can be used for drafting if your workplace rules permit it and you avoid sharing unnecessary personal or confidential information. Do not let the tool decide whether an issue is misconduct, poor performance or a formal employment matter.
How do I make AI feedback constructive?
Give it observable facts, the expected standard, the impact and a practical next step. Ask it to separate facts from interpretation, avoid comments on motives and include a question that gives the employee a chance to respond.
Should HR check AI-written employee feedback?
Ask HR or an experienced colleague to check feedback where the issue is disputed, repeated, sensitive or likely to affect formal performance action. A colleague can check clarity and fairness, but the manager still needs to verify the evidence and follow the employer's process.

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.