AI Feedback Prompts: Give Clearer Feedback at Work

Good feedback is specific, timely, and honest. Most managers give feedback that's vague, late, and hedged into uselessness. AI feedback prompts can help you get from "I have some concerns about Jake" to "Jake missed three consecutive deadlines on the Hendrix account, which pushed the client handoff back by two weeks." That's not magic. That's just having a thinking partner who doesn't get tired of helping you find the right words.

But let's be honest about what AI can and can't do here. It can sharpen language. It can separate your frustration from the actual facts. It can help you prepare for a hard conversation. It cannot give you courage. It cannot replace the relationship. And it absolutely cannot handle anything involving HR investigations, discrimination claims, harassment, or legal exposure. More on that at the end.

Here are 10 prompts you can use today.

The formula every AI feedback prompt needs

Before we get to the templates, you need to understand why most AI feedback attempts produce polished garbage. You paste in "my employee is always late and it's affecting morale" and get back a lovely paragraph about "punctuality expectations in a team context." Still useless. Just dressed up.

The fix is giving the AI actual inputs. Every prompt in this list follows the same structure:

Role + Specific behavior + Observable impact + What good looks like + Tone you want

Skip any of those and you get vague back. Feed them in and the output is actually usable. Think of it as Rule #13 from Don't Replace Me: garbage in, garbage out. Vague resentment becomes polished vague resentment. Specific examples become specific feedback.

Prompt 1: Turning messy notes into specific feedback

You've been scribbling notes for three weeks. You have: "missed thing again," "tone weird in that meeting," and "good job Friday." Not usable.

The prompt:

I'm a manager preparing written feedback for a direct report. Here are my raw notes from the past month: [paste your notes]. The employee's role is [job title]. Their main responsibilities include [list 2-3]. Please help me organize these into specific, observable feedback points. For each point, flag where I need to fill in a concrete example before this is ready to use.

The "flag where I need a concrete example" instruction is the key part. Don't skip it. You want AI to show you the gaps, not paper over them.

Prompt 2: Separating facts from interpretations

You think Jake doesn't care about the project. But "doesn't care" is an interpretation. What you actually observed was that he submitted two deliverables without spellchecking them and showed up to the client call without reviewing the brief. Those are facts.

The prompt:

I'm preparing feedback for an employee. Here's what I wrote: [paste your draft]. Please separate this into two columns: observable facts and behaviors, and my interpretations or inferences about motivation or attitude. Keep my exact wording for the facts column. Flag any interpretations that I'd need to rephrase before sharing them directly.

This one saves you from the conversation where the employee spends 20 minutes defending their attitude instead of acknowledging the behavior.

Prompt 3: Writing constructive criticism

The standard advice is "be direct but kind." Genuinely unhelpful. Direct and kind looks different depending on the person, the issue, and what actually needs to change.

The prompt:

I need to give an employee constructive feedback about [specific behavior or pattern]. The context is [2-3 sentences of background]. The impact of this behavior has been [what actually happened as a result]. The expectation for this role is [what good looks like]. Please draft feedback that is direct, specific, and focused on behavior and impact. Avoid softening language that dilutes the message. Keep it under 150 words.

Use this for written feedback or as a starting point for a verbal conversation. Always read it aloud before you use it. If it doesn't sound like you, edit it until it does.

Prompt 4: Writing specific praise

Generic praise is almost as useless as no praise. "Great work this quarter" tells someone nothing about what to repeat. Specific praise is a retention tool.

The prompt:

I want to recognize an employee for something they did well. Here's what happened: [describe the specific situation and what they did]. The result was [observable outcome]. Please write a short piece of specific positive feedback that names the behavior, explains why it mattered, and connects it to their role or team goals. Keep it genuine, not corporate.

Specific praise is also the part of feedback most AI handles well, since there's no conflict to manage. Still, read it. "I was genuinely impressed by your proactive stakeholder communication" sounds like a compliance form. Rewrite it in words you'd actually say.

Prompt 5: Preparing a feedback conversation

Most feedback goes sideways not because the words were wrong, but because the manager wasn't ready for the employee's reaction. Preparation is the job.

The prompt:

I'm preparing for a feedback conversation with an employee. The main issue I need to address is [describe the behavior and impact]. I've seen this pattern [number] times, most recently on [date or project]. The employee's role is [job title]. Please help me: (1) structure the conversation with an opening, the core feedback, and a path forward, (2) list 3 questions I should ask the employee to understand their perspective, and (3) identify one thing I might be missing that could change how I deliver this.

That third request is the valuable one. It forces the output to consider what you don't know.

Prompt 6: Role-playing the employee response

You say "the client presentation felt underprepared" and they say "I worked 60 hours on that." Now what?

The prompt:

I'm about to give feedback to [describe the employee in neutral terms: tenure, performance history if relevant, general disposition]. The feedback I'm planning to deliver is: [paste your draft]. Please play the role of this employee and give me 3 likely responses to this feedback, ranging from receptive to defensive to dismissive. For each response, suggest how I should continue the conversation.

This is one of the better uses of AI in feedback prep. You're not outsourcing judgment. You're war-gaming the conversation before it happens.

Prompt 7: Drafting a follow-up note

Feedback without follow-up disappears. A written note after the conversation isn't bureaucratic. It's a record of what was said and agreed.

The prompt:

I just had a feedback conversation with a direct report. Here's a summary of what we discussed: [key points, any commitments made, timeline]. Please draft a short follow-up note I can send within 24 hours that summarizes the conversation, confirms any agreed next steps, and keeps a constructive tone. Keep it under 200 words. Don't make it sound like an HR document.

That last instruction matters. A follow-up that reads like a disciplinary memo poisons the relationship even when the conversation went well. For more on drafting follow-up communications, the AI manager prompts guide has templates specifically for post-meeting notes.

Prompt 8: Turning feedback into a growth plan

The feedback conversation happened. Now what changes? A growth plan turns a single conversation into a direction.

The prompt:

Based on the following feedback I gave an employee: [paste feedback], and their role responsibilities [list them], please draft a simple 90-day growth plan. Include: 3 specific behaviors to develop, 2 ways we'll measure progress, and a checkpoint schedule. Keep it practical and achievable, not aspirational.

The checkpoint schedule is the part most growth plans skip. Without it, you check in at the end of 90 days and nobody remembers what the goal was. If you're working through a formal performance review process, the AI performance review prompts article has templates that connect directly to this kind of follow-through.

Prompt 9: Checking tone for clarity and respect

You wrote the feedback. You read it twice. You still can't tell if it sounds harsh or if you're just being precious about it.

The prompt:

Please review the following feedback draft for tone: [paste feedback]. Tell me: (1) Does any language come across as personal, dismissive, or attacking character rather than behavior? (2) Is anything unclear or open to misinterpretation? (3) Is there anything that softens the core message to the point where the employee might not grasp how serious this is? Please be direct. I'd rather know now than during the conversation.

That third check is the one most tone-review prompts miss. Feedback gets polished into meaninglessness as often as it stays too harsh.

Prompt 10: Deciding whether AI is the right tool at all

Some situations don't need better wording. They need HR, legal, or leadership.

The prompt:

I'm dealing with a workplace situation and I'm unsure whether to handle it myself, use AI to help me prepare, or escalate. Here's what happened: [describe the situation in general terms without names or identifiable information]. Please help me identify: (1) whether this situation has characteristics that typically require HR, legal, or leadership involvement, (2) what I might be overlooking, and (3) if this is appropriate for manager-level handling, what the first step should be.

This prompt is for you, not for AI to resolve the situation. It's a thinking framework.

Do not use AI tools to process, draft, or analyze situations involving: harassment claims, discrimination complaints, retaliation concerns, threats of violence, active HR investigations, protected employee health information, compensation data, or formal performance improvement plans. These require your HR team, employment counsel, or both.

The AI conflict resolution prompts article has more guidance on where that line sits.

What AI can't do in feedback situations

AI can make your language more precise. It can't make you more honest. It can't make you show up to the conversation instead of sending an email. It can't make an employee feel safe enough to actually hear you.

The managers who give the best feedback aren't the ones with the most polished language. They're the ones the employee believes. That's what Rule #4 in Don't Replace Me calls the trust currency. Feedback only lands when the person trusts your intent and trusts that your facts are real.

If you've used AI to replace the actual thinking, you'll notice. The employee will also notice. Something will feel slightly off about the specificity, or the phrasing won't sound like you, or the conversation will go sideways at the first pushback because you hadn't actually done the work of deciding what you believe.

Use these prompts to prepare. Then put the output away and talk to the person.

If you're just getting started with AI tools at work, the no-BS guide to using AI at work covers how to build a sustainable habit without the hype. And if you want to understand where AI genuinely breaks down before you rely on it for anything important, what AI can and can't do is worth five minutes.

The field guide for staying human when everyone else mistakes smoother wording for actual leadership is Don't Replace Me. The feedback chapter is the one most managers bookmark.

Frequently asked questions

Can AI write feedback for me?

AI can draft, structure, and refine feedback based on examples you provide, but it can't verify facts, understand your relationship with the employee, or substitute for your judgment. Always treat AI output as a first draft that needs your review before any conversation or documentation.

Is it okay to paste employee information into ChatGPT for feedback prep?

Only use general descriptions without names, identifiable details, compensation data, health information, HR investigation details, harassment or discrimination claims, or anything from formal performance management processes. When in doubt, ask your IT or HR team whether the tool is approved for this use.

What's the difference between feedback and a performance improvement plan?

Feedback is a normal part of managing: specific, timely, two-way. A performance improvement plan (PIP) is a formal HR process with legal implications and usually involves documentation, HR oversight, and defined consequences. Don't use AI to draft a PIP without HR involvement.

How do I give feedback when I don't have specific examples?

You don't. Vague feedback is worse than no feedback because it creates defensiveness without giving the person anything to act on. If you can't point to a specific behavior and its impact, you need to spend more time observing before having the conversation.

What should I do if the employee gets upset during the feedback conversation?

Pause. Acknowledge what you heard. Don't keep talking to fill the silence. Most feedback conversations go wrong because the manager keeps adding caveats when the employee gets emotional. Give them time to process. You can always continue the conversation at another time if the current one breaks down entirely.

When should I involve HR instead of handling feedback myself?

Involve HR when the situation involves harassment, discrimination, retaliation, a protected class, threats, formal disciplinary action, or anything that could have legal implications. Also involve HR if the employee has previously raised a complaint against you. When in doubt, a quick conversation with HR before the meeting is always better than a conversation with HR after things go wrong.