Home · Business · HR & People · Performance & training
As of 13 August 2026, AI can prepare for a difficult performance conversation.
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 comparable price is provided in the available tool data.
If this goes wrong: the conversation feels unfair or escalates a workplace dispute, and you still carry responsibility for the process and outcome.
What to actually do
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.
Use a tool built for this
Second choiceDo it yourself
The distant thirdA chat interface, chat-fluent skill, and roughly 30 minutes until you can act on the result.
How to actually do it
- Open your organisation's performance policy and the employee's role objectives, then copy only the relevant expectations and process steps into a separate working document.
- Gather dated, specific examples of the performance issue, including the expected standard, what happened, the work impact and any support already offered; remove names and unnecessary personal data.
- Write down the reasonable outcome you want from the meeting, such as an agreed improvement, support plan or follow-up date, without deciding the employee's motives.
- Paste the policy extracts, expectations, examples and desired outcome into the prompt, then ask the chatbot to produce the preparation brief and conversation plan.
- Compare every factual statement and proposed process step in the draft with your documents, and delete anything invented, personally judgemental or unsupported.
- Send the checked plan to your HR contact or a trusted manager for a fairness and process check before scheduling or holding the conversation.
- Take the final notes into the meeting, listen to the employee's response, and record the agreed actions, support and follow-up date in line with your organisation's policy.
Prompt
Help me prepare for a difficult performance conversation at work in the UK. I am the manager and the conversation concerns [role or situation]. The relevant expectations or policy are: [paste them]. The factual examples, with dates and impact, are: [paste only necessary work information and remove names or identifying details]. The outcome I need is: [describe the reasonable improvement or next step]. Create: 1. A short preparation brief separating confirmed facts, reported concerns and assumptions. 2. A calm opening that explains the purpose of the meeting without labelling the person. 3. A clear sequence for discussing each example, its impact and the expected improvement. 4. Open questions that give the employee a fair chance to respond. 5. Plain, respectful wording for likely defensive, upset or unexpected responses. 6. A closing that confirms agreed actions, support, timescales and the next check-in. 7. A list of statements I should not make because they are speculative, personal, discriminatory or unsupported. Do not invent facts, policy, dates or legal requirements. Do not diagnose a health condition or advise me to dismiss, discipline or discriminate against anyone. Flag any point that needs checking with my HR team or an employment solicitor. Keep the tone direct, private and respectful. Make the plan suitable for a real conversation, not a scripted speech.
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 evidence is complete, representative or being applied consistently across the team.
- AI cannot judge the employee's motives, personal circumstances or likely reaction from a written account.
- AI cannot take responsibility for the fairness of the process or the employment consequences of what you do next.
- AI cannot replace an HR check where the issue may involve sickness absence, disability, discrimination, protected characteristics or disciplinary action.
- AI cannot hold the conversation, listen to tone and body language, or adapt responsibly to facts that emerge in the room.
Even on a YES, the friction has a name: 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.
| Axis | Score (0–2) |
|---|---|
| Output | 2 |
| Inputs | 2 |
| Verification | 1 |
| Liability | 1 |
| Effort delta | 2 |
| Total | 8 / 10 |
The methodology and its thresholds are published in full.
FAQ
- Can AI help me prepare for a difficult performance conversation?
- Yes. It can turn your documented expectations and examples into a structured, respectful plan with questions, possible responses and agreed next steps. Check the facts and ask HR to review the process where the consequences are serious.
- Can AI write what I should say in a performance review?
- Yes, it can draft an opening, explanations of specific examples and a closing that sets out support and follow-up. Use it as notes rather than a script, and remove any wording that sounds personal, speculative or unsupported.
- What information should I give AI about an employee?
- Give it the role expectations, relevant policy wording, dated work examples, impact, support already offered and the outcome you are seeking. Remove names, health information and other identifying details unless your approved workplace system permits that use.
- Can AI tell me if a performance process is fair or legal?
- Not reliably. It can help you compare your draft with your organisation's policy, but it cannot take responsibility for UK employment decisions or assess all the relevant facts. Ask your HR team or an employment solicitor about a serious case.
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.