Use ChatGPT to turn approved performance notes into clear review language. Keep the evidence, evaluation, rating, and employment decisions with the manager and the organization’s review process.
The useful task is drafting and editing feedback you can substantiate. A fluent paragraph is not evidence that a person met expectations. This workflow starts with observed work and ends with a review you have checked yourself.
Prepare evidence before you ask for prose
Use the agreed role expectations and the full review period. For each point, record an observation, its source, the effect on work, and any relevant context. Separate documented facts, reported feedback, and your interpretation.
- Expected work: the responsibility or goal that was communicated.
- Observed work: a specific action, deliverable, or missed commitment.
- Impact: what changed for a customer, colleague, or project, with supporting evidence.
- Context: dependencies, resources, changed priorities, or support already discussed.
- Next step: an expectation, support commitment, and a date to review progress.
Do not infer motivation, health, or personality from a work outcome. Avoid treating visibility in meetings or time online as a substitute for performance evidence. Use the annual performance-review template to structure your notes.
Protect personnel information
Use only AI tools and workspaces approved by your organization for this purpose. Share the minimum approved information. Do not paste private personnel records, sensitive personal details, or raw peer messages into an unapproved tool. Removing a name does not necessarily anonymize a distinctive event.
The prompts below are for drafting from permitted notes. They do not authorize ChatGPT to rate employees, rank people, decide pay or promotion, or send a review. Apply your organization’s review and HR processes.
A copyable prompt for a review draft
Help me draft review language for [role] over [review period]. Use only these approved notes and role expectations: [notes]. Organize the output as contributions, development needs, and proposed next steps. For each point, distinguish the observation, impact, and next step. Use direct, respectful language under [word limit]. Do not invent examples, numbers, quotations, motives, or context. Mark missing evidence as [needs evidence]. Do not assign ratings or recommend pay, promotion, discipline, or termination.
The manager supplies and checks the evaluation; the model helps express it. This source-and-constraint approach follows OpenAI’s prompting guidance. Start with one review section rather than asking for every employee’s review at once.
Worked example: specific, balanced feedback
Illustrative notes: The role required a weekly release update by Tuesday noon. Over eight weeks, six updates arrived on time and two arrived Wednesday. Both late updates lacked an early warning. Partner teams had to change their planning. The employee’s technical explanations were clear. The manager will provide a shorter update template.
Possible draft: “Your release updates gave partner teams clear technical explanations. Six of the eight updates arrived by the agreed Tuesday deadline. The two late updates arrived Wednesday without an early warning, which meant partner teams had to replan. Next cycle, send the update by Tuesday noon or flag the risk on Monday with a revised delivery plan. I will provide a shorter template, and we will review the next four updates together.”
The draft recognizes a contribution, states the missed expectation, and includes support. It avoids labeling the person “unreliable” or guessing why the updates were late. Check that your own evidence supports the wording before using it.
Your rapid TeamDynamics · two minutes
What makes your team tick?
Start with how your meetings work. Answer 12 quick questions for a first look at how your team communicates, makes decisions, and gets things done.
When we have team meetings…
Question 1 of 12 · Choose an answer to startWrite development feedback the person can act on
Replace a trait label with a situation, behavior, effect, and request. “Needs executive presence” does not explain what to do. “Begin the steering update with the decision needed, then give the two main trade-offs” names a behavior to practice.
Review this paragraph for vague labels or judgments about the person: [draft]. Suggest wording that describes observed behavior, its work impact, and a specific request. Keep the severity and facts unchanged. If an example or expectation is missing, ask for it rather than inventing one.
Clear language should not soften a serious issue into vague encouragement. Use the normal organizational process for concerns requiring formal action; do not let a drafting tool decide the outcome.
Ten prompts for common review-writing tasks
- Organize notes: “Group these approved observations by the review sections. Keep source labels and dates. Do not evaluate or add facts.”
- Find unsupported claims: “For each factual claim in this draft, identify its supporting note or mark [needs evidence].”
- Clarify a contribution: “Rewrite this paragraph to distinguish the person’s contribution from the group result.”
- Make feedback specific: “Replace vague praise with the documented action and impact. Ask for evidence where it is missing.”
- Describe a missed expectation: “State the expected work, what happened, and its impact without guessing motives or changing the severity.”
- Check period coverage: “List the dates of the examples in these notes and highlight parts of the review period with little evidence.”
- Check wording consistency: “Compare these anonymized sample sentences against the same language checklist. Flag different standards of specificity; do not rank the people or suggest ratings.”
- Draft a next step: “Express this manager-selected development goal as an action, observable outcome, support commitment, and review date. Flag missing details.”
- Prepare discussion questions: “Suggest open questions that help me understand this work example without assuming why it happened.”
- Shorten the draft: “Remove repetition and jargon while preserving evidence, accountability, uncertainty, and the agreed next steps.”
Check fairness and accuracy yourself
- Use the same role-relevant expectations consistently; account for changes in scope or resources.
- Check every example, number, attribution, and quotation. A model’s self-check is not independent verification.
- Look for recency bias, personality labels, and unsubstantiated claims about intent.
- Make sure peer comments are represented faithfully and handled under the organization’s confidentiality rules.
- Read the wording against the evidence, not just for polish. Keep consequential judgments with the authorized human reviewers.
- Explain the feedback in a conversation, invite the employee’s perspective, and record agreed support and follow-up.
For broader input, use the 360-degree review template alongside your organization’s process. An employee preparing their own reflection can use the separate self-evaluation guide.
Separate team development from performance evaluation
A TeamDynamics team profile describes shared behaviors across Communicating, Processing, Deciding, and Executing. Personal preferences separately inform private CoDynamics; they do not determine the shared team result. Neither is a performance rating or a basis for pay, promotion, or other employment decisions.
Use team insights to discuss working agreements and support needs. For example, a team can clarify where decisions are recorded without labeling one member as the problem. Keep that development discussion distinct from the evidence used in an individual review.
