AI tools like ChatGPT can help with parts of performance management. But understanding what AI can and cannot do is essential to using it well. This guide includes 10 ready-to-use prompts, privacy guardrails, and an honest look at the limitations.
What AI Actually Helps With
ChatGPT is useful for tasks that benefit from language generation, structure, and outside perspective. It can help you:
These are language and structure problems. ChatGPT handles them reasonably well.
What AI Cannot Know
ChatGPT has no access to:
AI generates plausible-sounding text based on patterns. It doesn't know your team. Any output needs to be grounded in specifics you provide—and verified against reality before you use it.
Ready-to-Use Prompts
1. Preparing for a 1:1
I'm preparing for a 1:1 with a mid-level engineer who has been quiet in team meetings lately but is delivering solid work. I want to check in without making assumptions. Suggest 5 open-ended questions that create space for them to share what's on their mind.2. Giving Constructive Feedback
I need to give feedback to an engineer about missing deadlines. The pattern is: they commit to timelines in planning, then don't flag risks until the deadline passes. I want to address the behavior without being accusatory. Help me draft talking points.3. Writing a Performance Review Summary
Here are bullet points from my notes about an engineer's performance this cycle: [paste your notes]. Help me turn these into 2-3 paragraphs suitable for a written performance review. Maintain a balanced tone that acknowledges strengths and clearly states areas for growth.4. Discussing Career Growth
An engineer on my team wants to move into a tech lead role. They're strong technically but haven't demonstrated leadership behaviors yet. Help me outline a conversation that acknowledges their goal while being honest about what they need to develop.5. Reframing Harsh Feedback
I wrote this feedback and I think it's too harsh: "[paste draft]". Help me rewrite it to be direct but constructive. Don't soften the message—just improve the delivery.6. Responding to Defensive Reactions
I gave an engineer feedback about code quality issues. They responded defensively, saying the codebase was already messy when they joined. How do I acknowledge their point while still holding them accountable for their contributions?7. Structuring a Difficult Conversation
I need to tell an engineer they're not meeting expectations for their level. This is the first time I'm delivering this message formally. Give me a conversation structure: how to open, what to cover, and how to end with clear next steps.8. Generating Self-Review Prompts
I'm asking my team to write self-reviews. Many of them struggle with this. Give me 6 specific prompts I can share that help engineers reflect on their impact, growth, and challenges without just listing tasks completed.9. Calibrating Across the Team
I'm reviewing performance across 6 engineers before calibration. Help me create a simple framework for comparing contributions fairly, accounting for different project difficulties and role expectations.10. Identifying Blind Spots
Here's the performance review I drafted: [paste draft]. What might I be missing? Are there common biases I should check for? What questions should I ask myself before finalizing this?Guardrails for Privacy and Bias
Protect Your Engineer's Privacy
Before pasting anything into ChatGPT, consider what you're sharing. Avoid including:
- Full names or identifiable information
- Confidential project details
- Information about health, family, or personal circumstances
- Anything that would be inappropriate if leaked
Use generic descriptors: "a mid-level engineer" rather than names, "a critical infrastructure project" rather than specifics. ChatGPT doesn't need identifying details to help you structure feedback.
If your company has policies about AI tools and employee data, follow them. When in doubt, anonymize.
Watch for Bias in AI Output
ChatGPT reflects patterns in its training data, which includes societal biases. Be alert to:
- Gendered language: AI may use different descriptors for similar behaviors (e.g., "assertive" vs. "aggressive")
- Default assumptions: AI might assume certain roles, backgrounds, or career paths
- Recency in your inputs: If you only feed AI recent examples, it will reflect recency bias back to you
- Your own framing: AI amplifies the perspective you provide—if your input is skewed, the output will be too
Always review AI-generated text critically. Ask yourself: Would I write this about a different engineer in the same situation? Does this language match what I'd use for someone with a different background?
Don't Outsource Judgment
AI can help you phrase things. It cannot tell you whether your assessment is accurate. The hard work of performance management—observing, remembering, weighing context, making fair judgments—remains yours.
Where AI Falls Short
It Doesn't Know What Happened
ChatGPT will confidently generate performance feedback that sounds reasonable but has no connection to reality. If you ask it to write a review without providing detailed input, you'll get generic platitudes. The quality of output depends entirely on what you feed it.
It Can't Replace Observation
You can't prompt your way to good performance management if you haven't been paying attention. AI won't surface the contributions you didn't notice or remind you of feedback you forgot to give. It works with what you provide.
It Homogenizes Voice
AI-generated feedback tends toward a certain style: measured, professional, slightly generic. If all your reviews sound like ChatGPT wrote them, engineers notice. Your voice and specific observations matter.
It Doesn't Track Over Time
Each conversation with ChatGPT starts fresh. It doesn't remember that you discussed an engineer's communication issues three months ago, or that they've been working on a specific growth area. You need to provide that context every time—or maintain it elsewhere.
Preserving Context Over Time
One limitation of using ChatGPT for performance work is the lack of continuity. The tool doesn't maintain history across conversations, and reconstructing context every review cycle is tedious.
Some teams address this by maintaining structured notes in shared documents or dedicated tools. Vereda, for example, continuously captures context from 1:1s, standups, and check-ins throughout the review period—making it easier to surface patterns and evidence when review season arrives, rather than reconstructing from memory. The goal with any system is the same: ensure you're working from real data when it matters, not from what you can recall in the moment.
Using AI Responsibly
ChatGPT is a drafting tool, not a decision-making tool.
Use it to:
- Get unstuck when you're staring at a blank page
- Pressure-test feedback you've already drafted
- Find better words for things you already know
Don't use it to:
- Generate assessments about people you haven't observed
- Replace the work of paying attention throughout the year
- Avoid difficult conversations by hiding behind AI-generated language
The engineers on your team deserve feedback grounded in reality, delivered in your voice, based on evidence you actually have. AI can help you communicate more clearly. It cannot help you manage more fairly unless you do the underlying work.