Yes, for the draft. No, for the decision. AI drafts, the manager decides. Collect six factual inputs during the year, give them to an AI agent, and review the draft yourself before you share it. The scores, the judgment and the conversation stay with you.
What the research says
Managers already do this. In a survey by Highwire, a coaching programme of Knopman Marks, reported by HR Dive (8 October 2026), most US people managers said they had used AI tools for feedback or a review in the past 12 months. Far fewer employees were told.
| 78% | of 300+ US people managers used AI for employee feedback or a performance review in the past 12 months |
| 54% | of employees found AI-generated feedback more specific and actionable |
| 34% | said feedback has become more generic, and 32% said less useful |
| 16% | of 700+ individual contributors were told AI played a role in their most recent review |
Harvard Business Review (May 2026) puts the trade-off plainly: AI can make reviews faster and more polished, “but smoother language can also mask weak evidence, subjective judgments, and incomplete observations.” Its advice is to use AI to identify what the employee actually did: the decisions they made, the problems they solved, and the influence they had.
The Wall Street Journal (February 2026) frames it the same way in “The Do’s and Don’ts of Using AI to Write Performance Reviews”: it can help get your message across, “but it also could backfire badly.”
The process: six inputs, one draft, your review
A fair year-end review rests on facts collected during the year, not on what you remember in December. Gather six inputs, give them to an AI agent, and review the draft yourself. AI drafts. You make the assessment and the decision.
1. Yearly objectives, agreed earlier in the year
Start from the objectives you agreed with the team member earlier in the year. They should contain both work objectives and personal development items, so the person grows. For each objective, you assess whether it was reached on a 5-point scale:
| 1 | Not reached |
| 2 | To be improved |
| 3 | Achieved |
| 4 | Beyond expectations |
| 5 | Exceptional |
The score is yours. Set it before you open any AI tool.
2. Notes from every weekly 1-1
Add your notes from all the weekly 1-1 meetings of the year. They hold the concrete examples: what happened, when, and what was agreed. This is what HBR recommends basing the review on: the decisions the person made, the problems they solved and the influence they had. AI can also help you prepare each 1-1 during the year, so the notes exist when the review comes.
3. Your tone of voice and template format
Tell the AI how you write and what you want back: a one-page text or 4 slides. The draft should read like you, in the format your team member will receive.
4. Alignment with company strategy, if any
If your company has a strategy or yearly priorities, include them, so the review connects the person’s work to them.
5. Last year’s evaluation, if available
Add the prior year’s evaluation to track progress: which development points were addressed, and which remain open.
6. An agent that knows your role and your team’s roles
Give all of this to an AI agent that knows your manager role and the roles of your team members. Describe them once, in the agent’s instructions or at the start of the conversation, so the draft assesses the work against the right job.
Then: feed the inputs, get the draft, review it
Feed the AI all six inputs and let it produce a draft. Then review it yourself before you share it with the team member: check every statement against your notes, remove anything vague or unsupported, and confirm the scores are the ones you set. Tell the person how AI was used.
For the review, a second pass helps catch what you might miss. Then edit by hand and read it aloud as if the person were in front of you.
Risks to manage
- Bias. A model can repeat or polish biased wording that is already in your notes. The review check is there to catch personality labels and wording about the person instead of the work.
- Privacy. Performance notes are personal data. Use only a tool your company has approved for it, remove names of others, and keep health, salary and disciplinary details out of the prompt.
- Legal and HR policy. Check your company’s AI and HR policy first. In the EU, AI systems intended to monitor and evaluate the performance and behaviour of workers are listed as high-risk in Annex III of the AI Act. Drafting help is not a rating engine: keep the decision with a person. This guide is not legal advice.
- Employee trust. In the Highwire survey reported by HR Dive, only 16% of individual contributors were told AI played a role in their review. Highwire’s CEO, Liza Streiff, told HR Dive: “When employees suspect a tool shaped their review, and no one has told them how, playing it safe starts to feel like the smartest move.” Say how you used it.
As Streiff put it in the same report: “AI can help you write a faster review, but it doesn’t replace human judgment, context, nuance and empathy.” Put the time saved back into the conversation. Preparing that conversation is the subject of the week 3 note, One question, not a status round. The people side of the role is covered in What is an AI manager?
Try it before your next review
- Pick one team member and collect the six inputs.
- Score each objective yourself on the 1 to 5 scale.
- Run the prompt with an approved tool, then the review check.
- Edit by hand, and tell the person how AI was used before you share it.
Free kit
Get the free AI kit for managers: one-page tools you can use this week. AI prepares, you decide.
FAQ
Is it acceptable for a manager to use AI to write a performance review?
Yes, for the draft, if your company allows it. Give AI the facts and let it draft. The assessment, the decision and the conversation stay with the manager, and the employee should be told how AI was used.
What should I give the AI to draft a year-end review?
Six inputs: the yearly objectives (work and personal development) with your score for each on a 1 to 5 scale, your notes from the weekly 1-1s, your tone of voice and format (a one-page text or 4 slides), the company strategy if any, last year's evaluation if available, and a description of your role and the team member's role. Then review the draft yourself before you share it.
Should I tell employees that AI helped write their review?
Yes. In a Highwire survey reported by HR Dive in October 2026, 78% of managers said they had used AI for feedback or a review in the past 12 months, but only 16% of individual contributors said they were told AI played a role in their most recent review.
Can I paste performance notes into ChatGPT or another public AI tool?
Only into a tool your company has approved. Do not paste confidential data into a tool your company has not approved. Performance notes are personal data about an identifiable person: remove the names of others, and never paste health, salary or disciplinary details into an unapproved tool.
Can AI decide the rating?
No. The manager scores each objective on the 1 to 5 scale and makes the decision. AI turns those scores and the 1-1 notes into a draft. In the EU, AI systems intended to evaluate the performance of workers are listed as high-risk under Annex III of the AI Act.
Comments
No comments yet. Be the first to leave one.