How to use AI without outsourcing the courage in your performance conversation
“Karin, I hate writing. Is there a way I can use AI to help me document performance feedback?” #AskingForaFriend
Yes. And....
AI can help you prepare clearer, more specific, and more thoughtful feedback—in far less time.
And....it can also help you avoid thinking deeply, reinforce an unfair assumption, or bury an important message under three polished paragraphs of professional-sounding fog.
The question isn’t simply whether to use AI.
It’s which parts of performance management you can appropriately delegate—and which parts still belong to you.
Here’s the simplest rule I know:
Delegate preparation to AI. Don’t delegate your judgment, the relationship, or the conversation.
A Human-Centered Workflow for Using AI in Performance Feedback
AI is most useful when you give it a clearly defined role at each stage of the process.
1. Use AI to Organize What You Know—Not Invent What You Missed
You’re preparing for a performance conversation, but your notes are scattered across emails, project documents, meeting agendas, and whatever you scribbled down after that difficult customer call.
AI can help you organize information you have already documented.
For example, after removing identifying and confidential information, you might ask:
“Organize these project notes chronologically. Group them into accomplishments, observable behaviors, customer impact, collaboration, missed commitments, and agreed-upon next steps. Don’t add, infer, or interpret information.”
That can save you time and make patterns easier to see.
But AI cannot tell you whether your notes are complete, representative, or fair.
Perhaps you documented every mistake but neglected to record the employee’s quieter contributions. Maybe priorities changed halfway through the year. Perhaps the employee was working around an organizational obstacle you never removed.
Use AI to organize the evidence. Don’t ask it to manufacture the evidence—or decide what the evidence means.
And before entering anything, follow your organization’s policies. Don’t put employee names, medical information, accommodation requests, investigation details, compensation information, or other confidential material into an unapproved tool.
See content credentials
Click here to learn about our open-enrollment Human-Centered Peformance Management Live-Online Training
2. Use AI to Separate Observations From Assumptions—Not Diagnose the Employee
This may be one of the most valuable ways AI can help.
Performance feedback often starts with a legitimate concern but quickly drifts into interpretation:
- “She doesn’t care about the team.”
- “He lacks initiative.”
- “They aren’t leadership material.”
- “She has a bad attitude.”
- “He’s resistant to change.”
Those statements might reflect a real problem. But they don’t describe what happened.
Try asking:
“Review these de-identified observations. Separate specific, observable behavior from my interpretations about attitude, personality, motivation, or intent. Identify the conclusions I need to examine or discuss with the employee.”
You might discover that:
- “Doesn’t care about the team” is an interpretation.
- “Didn’t respond to three requests for a project update” is an observable behavior.
- “Isn’t leadership material” is a conclusion.
- “Interrupted colleagues six times during the project review” is an observable behavior.
Now you have something you can discuss.
What AI should not do is determine whether someone is committed, disengaged, ambitious, difficult, trustworthy, or ready for leadership.
AI can generate an explanation. That doesn’t mean it has discovered the truth.
You learn what’s happening through observation and conversation—not an AI-generated diagnosis.
3. Use AI to Make Your Feedback Clearer—Not More Comfortable
This is where many managers get into trouble.
They begin with feedback that feels too direct and ask:
“Make this sound nicer.”
AI obliges.
You start with:
“Jordan raised significant project risks too late for the operations team to respond.”
AI gives you:
“Jordan consistently contributes valuable perspectives and demonstrates a strong commitment to team success. Continued focus on proactive communication and strategic ownership will help elevate Jordan’s overall impact.”
It sounds polished.
It also tells Jordan almost nothing.
I call this Diaper Genie feedback—wrapping the real issue in so many layers of pleasant language that no one can tell what’s inside.
If you're new to my diaper genie metaphor, I explain why Diaper Genies don’t belong at work—particularly when you’re giving feedback or avoiding a conversation you know you need to have.
AI didn’t create Diaper Genie feedback. It simply made it easier to produce.
Instead of asking AI to make your message “nicer,” try:
“Identify vague phrases, corporate jargon, generalizations, and places where I have softened the central message. Show me where I need a specific behavior, example, impact, or expectation. Don’t invent those details for me.”
Once you’ve added the missing specifics, you can ask:
“Help me express this in clear, respectful, conversational language. Preserve the observable behavior, its impact, and the expectation going forward. Don’t obscure or minimize the concern.”
Your goal isn’t to make important feedback pleasant.
Your goal is to make it clear, respectful, and useful.
4. Use AI to Prepare Better Questions—Not Script the Entire Conversation
A performance conversation should be a conversation.
AI can help you identify questions that invite the employee’s perspective:
“Based on this de-identified situation, suggest five open-ended questions that will help me understand the employee’s perspective. Avoid questions that assume blame, motivation, or intent.”
You might prepare questions such as:
- "How does thsi look from your perspective?”
- “I'm curious made it difficult to raise this concern sooner?”
- “What information or support would have helped?”
- “How would you approach this differently next time?”
- “Is there context I may be missing?”
AI can also help you anticipate where the employee might need greater clarity.
But don’t ask it to predict exactly how the person will react or write both sides of the conversation. And don’t become so attached to your AI-generated script that you stop listening.
The employee may disagree, share information you didn’t know, or help you see the situation differently.
That isn’t the conversation going off track.
That is the conversation.
5. Use AI to Summarize What You Agreed—Not Decide the Outcome in Advance
After the conversation, an approved AI tool can help turn your notes into a concise recap.
You might ask:
“Organize these de-identified notes into the expectation discussed, the employee’s perspective, the agreed-upon action, the support I will provide, how success will be measured, and the follow-up date. Don’t add commitments or attribute statements that aren’t in my notes.”
Then review every word.
A useful follow-up might begin:
“Here’s what I heard us agree to today. Please let me know if I missed or misunderstood anything.”
What you shouldn’t do is create the entire summary before the conversation and merely fill in the employee’s name afterward.
The employee’s perspective should influence the next steps. Otherwise, you’re using the appearance of dialogue to deliver a predetermined outcome.
Four Things That Should Still Belong to You
Regardless of how capable AI becomes, managers should not delegate:
- The judgment: Don’t ask a general-purpose AI tool to assign performance ratings or make promotion, disciplinary, or termination decisions.
- The interpretation: Don’t ask AI to determine attitude, motivation, potential, commitment, personality, or the cause of poor performance.
- The responsibility: Don’t treat an AI-generated recommendation as an objective answer. You and your organization still own the decision.
- The conversation: Don’t send an AI-generated message to avoid a discussion that deserves your presence and attention.
AI can help you prepare what to say.
It cannot notice the hesitation before an employee answers. It cannot respond thoughtfully when they share unexpected context. It cannot repair trust, recognize courage, or demonstrate that you care about their success.
That’s the work of leadership.
Before You Use AI-Assisted Feedback, Ask Yourself
- Can I personally verify every example?
- Have I separated observable behavior from my assumptions?
- Is the impact clear?
- Does the employee know what needs to happen next?
- Have I made room for their perspective?
- Would I say these words out loud?
- Am I prepared to listen if they disagree?
If the answer is no, you’re not ready to hand the draft to AI.
You have a little more leadership work to do first.
AI can help you organize the facts, pressure-test your assumptions, clarify your message, prepare thoughtful questions, and document what you agreed.
Those are meaningful benefits.
Use them.
Then close the tool, look your employee in the eye, stay curious, and have the conversation.
