AI for managers: one-to-ones, feedback and reviews
Use AI to prepare one-to-ones, structure feedback and word hard messages, without handing it your team’s private details or your judgement about people.
Video transcript
One-to-ones, feedback, objectives and awkward messages: AI can help you find the words. Here’s how to use it without giving away your team’s privacy, or your judgement.
Start with what goes in. Work from your own notes, with names taken out, in a tool your organisation approves. Keep health details, HR cases, pay and grievances out.
Before a one-to-one, paste in your notes and ask for an agenda, some open questions, and anything you promised to follow up last time.
For feedback, try situation, behaviour, impact. Say when and where it happened, what the person actually did, and the effect it had. AI is good at reshaping vague notes into that shape.
But don’t ask it to judge anyone’s performance. It only knows what you typed, it tends to agree with you, and it can repeat biases from the writing it learned from.
It’s also a quick first drafter for objectives, role descriptions and difficult messages. Write the facts yourself, then ask it to make the message clear and kind.
And be open. Tell your team what AI helps you with, and that every decision about their work is still yours. Follow your organisation’s AI policy, too.
Copy the prompts from the full guide below, and try one before your next one-to-one.
In 30 seconds
- Work from your own notes, with names and personal details taken out, in a tool your organisation approves.
- Use it to structure and word things: feedback, objectives, role descriptions and difficult messages.
- Judging people stays with you. Watch for bias, and tell your team how you use AI.
It’s Sunday evening. You have five one-to-ones tomorrow, and your notes are scattered across a notebook, your inbox and a sticky note. AI can turn them into a plan and help you find the words for harder conversations. What it shouldn’t do is make up its mind about your people.
What goes in, and what stays out
Information about your team, such as performance concerns, sickness, pay or a grievance, is personal data under Jargon busterUK GDPR: The UK law on personal data. It sets how organisations must collect, use and protect information about people., and some of it, such as health, needs extra care. Keep it out of AI tools unless your organisation has approved that use, and follow its AI policy. No policy yet? Here’s what a good one covers.
Even in an approved tool, work from your own notes rather than their records, and take names out. Then write as if they’ll read it, because they might: staff can make a Jargon busterSubject access request: A request to an organisation for a copy of the personal information it holds about you. UK GDPR gives people the right to make one. for the personal information held about them, and that can include appraisal notes and emails.
Prepare one-to-ones from your own notes
Paste in what you’ve jotted down since last time: what went well, what’s stuck and what you promised to do. Ask for an agenda, and for open questions that let them do most of the talking.
I’m preparing a [30]-minute one-to-one with a [role] in my team. Here are my notes since our last meeting, with names removed: [your notes]. Suggest a short agenda, five open questions that let them do most of the talking, and a list of anything I promised to follow up. Don’t judge their performance: just organise my notes.
Feedback that’s specific and fair
Vague feedback, such as “be more proactive”, is hard to act on. The Jargon busterSBI model: A way to structure feedback in three parts: the situation, the person’s behaviour and its impact. It keeps feedback about actions, not personality. gives it a shape: the situation, what the person did, and the impact it had. It works for praise as well as problems, and for the evidence in a performance review.
| Instead of | Try |
|---|---|
| “You’re not a team player.” | “In Tuesday’s planning meeting, you talked over Sam twice, so we didn’t hear his idea on the budget.” |
| “Great job on the report.” | “Putting the figures first in your board report meant the board decided in ten minutes.” |
| “You need to be more proactive.” | “When the supplier missed their date, you waited to be asked. Flagging it that day would have given us time to find another.” |
AI is good at this reshaping. Give it what you saw, and ask it to separate behaviour from your interpretation. Beware Jargon busterSycophancy: When an AI tells you what you want to hear, agreeing with you instead of giving an honest view., though: if you go in annoyed, it will happily help you build a case. And never ask it whether someone is performing well. It only knows what you typed.
Help me turn these notes into feedback using situation, behaviour, impact. Notes: [what you saw, with names removed]. Keep to what I observed. Point out anything in my notes that is an assumption, or a judgement about personality rather than behaviour, and suggest a question I could ask instead.
Objectives, role descriptions and hard messages
AI is a quick first drafter. Ask for objectives that are specific and measurable, then agree them with the person, not just with yourself. For role descriptions and job adverts, ask it to flag wording that could put people off, and requirements the job doesn’t need, such as a degree for a role that doesn’t use one.
For difficult messages, such as turning down a request or announcing a change, write the facts yourself first. Then ask AI to make the message clear and kind, without softening what’s been decided.
I need to tell my team about [the change, such as new shift patterns] from [date]. What’s decided: [the facts]. What’s still open: [what you don’t know yet]. Draft a short, honest message that leads with the change, explains why, and says what happens next and who to ask. Keep it warm but clear, and don’t promise anything I haven’t listed.
Be open about how you use it
Tell your team what AI helps you with, and what it doesn’t: for example, it organises your notes and suggests wording, but every judgement about their work is yours. Acas, the workplace advice service, recommends talking to staff about AI early, and checking what it produces for accuracy, tone and bias.
Then hold yourself to the standard you’d expect of them: approved tools, no personal details, and a person making every decision about people. For the basics, see using AI at work safely.
Check yourself
3 quick questions nothing is savedSources (4)
- Subject access request Q and As for employersInformation Commissioner’s Office
- “Kelly is a Warm Person, Joseph is a Role Model”: Gender Biases in LLM-Generated Reference LettersWan and others, Findings of EMNLP 2023, December 2023
- Protected characteristicsEquality and Human Rights Commission
- One third of employers think AI will increase productivityAcas, May 2025
Spotted a mistake? Tell us and an editor will check it.
