Write Performance Reviews with AI

1. 1

   Gather Evidence and Raw Notes

   Performance reviews fail when they're based on recent memory and vague impressions. The first step is collecting concrete evidence — specific examples, metrics, and incidents from the full review period.

   ChatGPT Claude

   I need to write a performance review and want to make sure I'm capturing the right evidence. Help me run a structured evidence-gathering exercise.
   
   Context:
   - I am: [a manager writing a review for a direct report / an employee writing a self-review]
   - Employee name and role: [Name, e.g., 'Sarah, Senior Software Engineer']
   - Review period: [e.g., 'January - December 2025']
   - Review purpose: [annual review / mid-year check-in / promotion review / PIP check-in]
   
   Raw notes and examples I can recall:
   [Dump everything you remember — unfiltered. Projects they worked on, things that went well, things that didn't, feedback you gave, metrics, incidents, moments you noticed their growth, moments that were frustrating. Don't organize it. Just write it all down. Example: 'Shipped the API refactor project 2 weeks late but quality was excellent. Got great feedback from the product team. Was slow to ask for help when blocked on the auth issue in March. Mentored the new intern well — James said she was the most helpful person on the team. Missed two sprint goals in Q2 due to scope creep she didn't push back on. Presented to senior leadership in November and did well.']
   
   Given this raw input, help me:
   
   1. **Identify specific evidence** for each of these categories: technical/functional performance, collaboration, communication, ownership and initiative, growth and learning, impact on team/company.
   
   2. **Flag what's missing**: What evidence gaps do I have? What questions should I be asking myself to fill them? (E.g., 'Do you have any metrics on the projects she delivered? What feedback did you receive from stakeholders?')
   
   3. **Separate facts from interpretations**: Go through my notes and flag which statements are observable facts vs. my interpretation. For example, 'was slow to ask for help' is an interpretation — the fact might be 'blocked on auth issue for 3 weeks before escalating.'
   
   4. **Rate the evidence quality**: Which claims do I have strong evidence for? Which are based on impressions that could be challenged? This helps me write only what I can actually support.

   Copy

   Tip: Keep a running document of notes throughout the year — 'the manager's field notes.' One sentence per observation, dated. When review season arrives, you're pulling from 12 months of documented examples instead of trying to remember what happened in February. Fifteen seconds of documentation per week saves 3 hours of struggling during review season.

2. 2

   Write the Core Performance Narrative

   Turn your evidence into a coherent narrative that's specific enough to be meaningful, fair enough to be defensible, and actionable enough to actually guide development.

   ChatGPT Claude

   Help me write the core performance narrative for [employee name]'s [review period] performance review.
   
   Role: [employee's title]
   Overall performance level: [e.g., 'Meets expectations' / 'Exceeds expectations' / 'High performer' / 'Needs improvement']
   
   Key evidence (from Step 1):
   - Achievements: [list 3-5 specific accomplishments with context]
   - Strengths demonstrated: [list 2-3 with examples]
   - Areas for growth: [list 1-3 with specific examples, not general criticisms]
   - Impact on team/company: [specific contributions, metrics if available]
   
   Company review template sections I need to fill:
   [If you have a template, list the section headings. Otherwise, use: Summary, Key Accomplishments, Strengths, Development Areas, Goals for Next Period]
   
   Write each section with these standards:
   
   1. **Specific over general**: Replace 'does good work' with 'delivered the authentication module on schedule with zero post-launch bugs, despite a mid-project scope change that required redesigning the permission system'
   
   2. **Evidence-anchored**: Every claim should be traceable to a specific project, incident, or pattern of behavior — not 'often struggles with communication' but 'in three separate stakeholder updates (March, June, October), technical details required follow-up clarification from the product team'
   
   3. **Balanced and fair**: Achievements section should celebrate real wins. Development section should identify real growth areas without softening them to uselessness — honest, specific, and constructive
   
   4. **Future-focused development areas**: Frame each growth area as a skill to build, not a criticism: 'opportunity to strengthen proactive communication when blocked on dependencies' not 'tends to go silent when stuck'
   
   5. **Tone**: Professional, direct, and respectful. Avoid both empty praise ('great team player!') and harsh judgment ('lacks initiative'). Sound like a manager who actually knows and cares about this person.

   Copy

   Tip: The development section is where most managers write the most useless content. 'Needs to improve communication' tells the employee nothing. 'When blocked on cross-team dependencies, tends to wait rather than escalate — in Q2 this delayed the checkout feature by two weeks. Work on identifying blockers earlier and coming to 1:1s with a specific ask' is actionable.

3. 3

   Write SMART Goals for the Next Period

   Goals that are vague get ignored. Write specific, measurable goals tied to the employee's development areas and the team's direction.

   ChatGPT Claude

   Help me write SMART goals for [employee name]'s next review period.
   
   Development areas we identified: [list from Step 2]
   Team priorities for next period: [what the team will be focused on in the next 6-12 months]
   Employee's career aspirations (if known): [e.g., 'wants to move into a tech lead role' / 'interested in management' / 'wants to deepen technical expertise in ML']
   Role expectations that aren't fully met yet: [if any]
   
   For each goal, write using the SMART framework:
   - **Specific**: What exactly needs to happen?
   - **Measurable**: How will we know it's been achieved? (Quantitative or clear binary outcome)
   - **Achievable**: Is this realistic given their current level and the timeframe?
   - **Relevant**: Does this connect to their development AND the team's needs?
   - **Time-bound**: By when? With what interim checkpoints?
   
   Create:
   1. **2-3 performance goals** (tied to the role's core responsibilities and team outcomes)
   2. **1-2 development goals** (tied to skills they're building toward the next level)
   3. **1 stretch goal** (ambitious but achievable — something that would represent genuine growth)
   
   For each goal include:
   - The goal statement (1-2 sentences)
   - How it will be measured
   - What support or resources they need from me as the manager
   - 90-day checkpoint milestone

   Copy

   Tip: The best goals are ones the employee helped write. Before finalizing, share your draft goals in the review meeting and ask 'does this feel fair and achievable? Is there anything you'd add or adjust?' Goals that the employee co-created get dramatically more follow-through than goals handed down from above.

4. 4

   Edit for Tone, Fairness, and Legal Safety

   Final review to catch language that's vague, biased, legally risky, or that would embarrass you if read aloud in a meeting.

   Claude Grammarly

   Review this performance review draft for quality, fairness, and legal/HR safety before I submit it.
   
   [Paste your complete draft]
   
   Employee's role: [title]
   Overall rating: [rating]
   
   Check for:
   
   1. **Specificity audit**: Flag every claim that isn't backed by a specific example or data point. Replace vague language with specific language suggestions. E.g., 'hard worker' → needs a specific example; 'good communicator' → needs a specific instance.
   
   2. **Bias check**: Flag any language that might reflect:
      - Recency bias (are the examples all from the last 2 months?)
      - Halo/horn effect (is the entire review colored by one exceptional or one poor incident?)
      - Personality vs. performance confusion (are you evaluating how much you like this person vs. their work outcomes?)
      - Protected class language (any mention or implication of age, gender, family status, health, etc.)
   
   3. **Legal risk flags**: Flag any language that could be problematic if this review was reviewed by HR, an employment lawyer, or used in a termination or legal proceeding. Common issues: subjective personality assessments, references to personal life, promises about future compensation, ambiguous ratings on a documented PIP.
   
   4. **Balance check**: Count the words spent on achievements vs. development areas. Is this proportional to the overall rating? A 'meets expectations' review should have roughly 60% achievement, 40% development. A 'needs improvement' review can flip that.
   
   5. **Actionability**: Will the employee leave the review knowing exactly what to do differently? Flag any development feedback that's too vague to act on and suggest more specific language.

   Copy

   Tip: Read the review imagining the employee reads it for the first time in their car after the meeting. Will they feel fairly seen? Will they know what to do next? Will anything feel like a surprise they've never heard before? If you can answer yes to all three, the review is ready to submit.

Recommended Tools for This Scenario

ChatGPT

Freemium

The AI assistant that started the generative AI revolution

• GPT-4o multimodal model with text, vision, and audio
• DALL-E 3 image generation
• Code Interpreter for data analysis and visualization

Get Started →

Claude

Freemium

Anthropic's AI assistant built for thoughtful analysis and safe, nuanced conversations

• 200K token context window for massive document processing
• Artifacts — interactive side-panel for code, docs, and visualizations
• Projects with persistent context and custom instructions

Get Started →

Grammarly

Freemium

AI-powered writing assistant for grammar, clarity, and tone

• Real-time grammar, spelling, and punctuation correction
• Tone detection and adjustment
• Full-sentence rewrite suggestions

Get Started →