The promise of a Microsoft Copilot performance review is simple: less blank-page time, more useful feedback. A manager with 12 direct reports can spend a full week drafting review forms, chasing peer comments, and rewriting the same phrases. Copilot can absorb some of that grunt work. It cannot decide whether someone deserves a promotion, and it should not be trusted to invent examples. The useful split is clear: Copilot handles structure, summaries, and tone. You handle evidence, judgment, and the conversation.
Here is a practical workflow for using Microsoft Copilot in a performance review cycle, including prompts, privacy checks, and ways to keep the feedback specific.
What Microsoft Copilot can actually do in a performance review
Copilot inside Microsoft 365 can work with content you already have access to, such as emails, Teams messages, and documents. In a review cycle, that means it can:
- Turn six peer comments into three recurring themes.
- Draft a first version of a review from your notes.
- Rewrite blunt feedback into clear, professional language.
- Create a conversation guide with questions and talking points.
- Suggest next-quarter goals based on the employee’s role and previous objectives.
What it cannot do is observe the employee’s actual work. If your notes are thin, Copilot will produce polished generalities. If your notes are biased toward the last three weeks, the draft will inherit that bias. Treat it like a fast writing assistant, not a performance judge.
It also matters which Copilot you use. Microsoft 365 Copilot works inside Word, Outlook, and Teams with your organization’s permissions. Copilot Chat is a more conversational tool, and its data handling depends on the account and license. If you are unsure what your version includes, this breakdown of how Copilot Chat works, what it costs, and who is using it is a useful starting point. For hands-on practice, this step-by-step Copilot Chat walkthrough with real work examples shows how to structure prompts without overcomplicating them.
A simple workflow from raw notes to review draft
1. Gather evidence before you open Copilot
Write down three to five concrete examples per employee. Each one should include the situation, the behavior, and the impact. “Handled the billing migration in March” is a start. “On March 14, flagged a duplicate-charge issue in the billing migration, which prevented about 400 customers from receiving incorrect invoices” gives Copilot something to work with. Add peer feedback, goal results, and relevant dates. A simple note file is enough. Copilot does not need a perfect dataset, but it does need specifics.
2. Ask for structure, not verdicts
Vague prompts produce vague reviews. Instead of “Write my performance review for Alex,” try a prompt that defines the sections and the evidence. For example:
Prompt: “Using the notes below, draft a performance review with three sections: strengths, growth areas, and two goals for next quarter. Keep the tone direct and professional. Use only the examples provided. Do not add achievements or metrics that are not in the notes. Each strength should cite one specific example.”
That last instruction matters. Copilot is good at filling gaps, which is exactly what you do not want in a formal review. A second prompt can tighten the language: “Rewrite this section for clarity. Remove generic praise. Keep the specific examples and dates.”
3. Edit for specificity and fairness
Read the draft with a red pen, even if it sounds good. Replace phrases like “strong team player” with the behavior you actually saw. Check whether you used the same standards for everyone. A useful test is to line up five reviews and scan for patterns. If one employee’s draft is full of “needs to improve communication” and another’s is full of “led cross-functional work,” ask whether the difference reflects performance or the way you wrote the notes. Copilot mirrors your inputs. It will not correct a manager who only documented certain kinds of work.
4. Prepare the conversation, not just the document
The review form is half the job. The meeting is where the feedback lands. Ask Copilot to turn the draft into a 30-minute conversation plan: five minutes for outcomes, ten minutes for examples, ten minutes for the employee’s perspective, five minutes for goals. It can also generate questions such as “Which part of this feedback is most surprising?” or “What support would make the next quarter easier?” Practice the opening out loud. A review that reads well can still sound stiff when spoken.
Prompts that work better than “write my review”
Good prompts narrow the task and set boundaries. These four patterns cover most of the cycle:
- Summarize themes: “Group these eight peer comments into three themes. Quote one phrase from each comment as evidence. Flag any contradictions.”
- Draft with constraints: “Write a 250-word review using only the notes below. Use plain language. Include one growth area and two measurable goals.”
- Remove bias traps: “Review this draft for vague language, personality judgments, and recency bias. Suggest specific replacements.”
- Build the meeting plan: “Create a 30-minute agenda for a performance review conversation based on this draft. Include three open-ended questions.”
Keep a prompt library in a shared team document. Managers who reuse a tested structure get more consistent results than managers who improvise a new prompt for every employee.
Where Copilot helps most, and where it gets in the way
Copilot earns its keep during the messy middle of the process. Summarizing peer feedback, turning bullet points into paragraphs, and drafting goal statements are all low-risk tasks with a clear source. It also helps managers who tend to write either too bluntly or too vaguely. A prompt like “make this direct but not harsh” gives them a second pair of eyes before the conversation.
The problems show up when managers skip the evidence step. Copilot will happily produce a review that sounds complete but contains no real observation. It may also over-polish. A sentence like “Alex leveraged cross-functional synergies” sounds like a press release, not feedback. Cut it. Employees remember specific moments, not adjectives.
There is a process angle too. If your organization is redesigning review workflows around AI agents, the lessons in this piece on enabling agent-first process redesign apply: start with the decision you need to make, then assign the assistant a bounded task. Do not automate the entire review cycle and hope for the best.
Privacy, bias, and the paper trail
Performance reviews contain sensitive personal data. Use your work account and the company-approved Copilot, not a personal chatbot. Check what your HR and legal teams allow before pasting in peer comments, medical information, or disciplinary notes. A good rule is to minimize data: give Copilot the examples it needs, not the entire employee file.
Bias deserves its own check. Language models pick up patterns in the text you provide. If your notes focus on how someone speaks in meetings and ignore their delivered work, the draft will do the same. Before you finalize, ask:
- Are the examples tied to outcomes, not personality?
- Would I write this sentence about another team member?
- Did I include recent work only, or the full review period?
- Are the goals achievable with the resources the employee actually has?
If you are comparing assistants for drafting, this comparison of ChatGPT vs Claude is helpful for general writing work. Copilot’s edge in a Microsoft shop is context: it can work inside the tools where the review evidence already lives. That advantage disappears if managers export everything into a side tool with weaker controls.
How to measure whether it saved time
Do not rely on a feeling that Copilot “felt faster.” Track three numbers before and after a pilot: minutes spent preparing each review, number of revision rounds, and manager confidence in the final feedback. Run the pilot with eight to ten managers for one cycle. If preparation drops from 45 minutes to 25 minutes per review, a 10-person team saves about 3.3 hours across the cycle. That is useful, but only if feedback quality holds up.
Add a simple quality check. Ask employees whether the review included at least two specific examples and whether the goals felt relevant. Compare the results with the previous cycle. If the drafts are faster but more generic, the workflow needs work. If managers are spending less time on formatting and more time on evidence, you are using the tool well.
A practical test for your next review cycle
Pick five employees and use Copilot for preparation only. Gather notes, ask for a structured draft, edit it yourself, and run the conversation without reading a Copilot-generated script. After the meetings, note which prompts saved real time and which produced filler. Then decide whether to expand the workflow. The best Microsoft Copilot performance review process is not the one with the most automation. It is the one that gets managers to collect better evidence, write clearer feedback, and walk into the room prepared to listen.

