1. Home
  2. Prompt library
  3. Hiring and management
Prompt library

Hiring and management prompts

Hiring and managing people is where AI helps with the writing and the structure while you keep every judgment. It can turn a rough list into a job description, build a scorecard so every candidate gets the same questions, and draft a performance review from your notes that sticks to observed behavior. What it cannot do is decide who to hire, promote, or let go, and it should never be the reason a decision was made.

Three rules apply to every prompt here. Remove names and identifying details about candidates and employees before pasting anything into a consumer AI tool unless your organization has approved it for HR data. Watch the output for bias: the AI learned from a lot of hiring language that quietly favors some groups, so these prompts ask it to stick to job-related criteria and flag anything touching a protected characteristic (race, sex, age, disability, religion, national origin, pregnancy, and others under EEOC rules and your state's laws). And some states and cities regulate automated tools in hiring decisions, so keep a human making every decision and check your jurisdiction.

Write a job description from a rough list of duties

You know what the job is but what you have is a bulleted list and a title. Use this to get a description that attracts the right people and does not scare off good candidates with a wish list.

Act as an experienced recruiter who writes job descriptions that attract strong, diverse candidates and screen out the wrong ones honestly.

Job title: [JOB TITLE]
My rough list of duties, requirements, and thoughts: [PASTE YOUR ROUGH LIST]
Team, manager, and how the role fits: [TEAM AND REPORTING LINE]
Location, remote or hybrid policy, and schedule: [LOCATION AND WORK ARRANGEMENT]
Pay range and benefits I will publish: [PAY RANGE AND BENEFITS]
Our company in one or two sentences: [COMPANY DESCRIPTION]

Ask me anything you cannot infer, especially what success looks like in the first year. Then write:
1. A two-sentence summary of why this role exists.
2. What you will do: five to seven outcomes, written as results rather than tasks.
3. What you need: only true must-haves, each tied to a duty above. Put everything else under 'nice to have' or cut it. Prefer skills and experience over years and degrees unless a credential is legally required.
4. Pay, benefits, location, and how to apply.

Use plain, inclusive language. Avoid gendered or age-coded words (rockstar, ninja, digital native, young and energetic) and anything referencing a protected characteristic. Flag any requirement from my list that could unintentionally exclude qualified people, and tell me why.

Tip: Many states and cities now require a pay range in postings; check the rule for every location you hire in. And read the must-haves list twice: every item you keep will cost you candidates, so keep only what the job truly needs.

Works well in: ChatGPT, Claude, Gemini, Microsoft Copilot

Build a structured interview scorecard

You want every candidate asked the same questions and rated against the same standard, so the decision rests on evidence instead of who the interviewer liked. Use this before the first interview, not after.

Act as a talent assessment specialist who designs structured interviews that are fair, consistent, and legally sound.

Job title: [JOB TITLE]
Job description: [PASTE THE JOB DESCRIPTION]
The five competencies that matter most for success in this role: [TOP 5 COMPETENCIES]
Interview length and format: [INTERVIEW LENGTH AND FORMAT]
Who is interviewing and their experience with structured interviews: [INTERVIEWERS]

Build a scorecard that includes:
1. For each competency: a one-line definition of what it looks like in this job, two behavioral questions about past experience ('Tell me about a time...'), and two follow-up probes to get specifics on the situation, the candidate's own actions, and the result.
2. A four-point rating scale for each competency with observable anchors: what a 1, 2, 3, and 4 answer actually contains. No personality adjectives.
3. A note-taking template that separates what the candidate said from the interviewer's interpretation.
4. A short list of topics interviewers must not ask about or factor in, including anything touching age, family or pregnancy plans, health or disability, religion, national origin, citizenship beyond legal work authorization, or salary history where prohibited.
5. Instructions for scoring independently before any group discussion.

Keep the questions the same for every candidate.

Tip: Have every interviewer submit scores before the debrief. The moment one person says 'I loved her', everyone else's ratings drift, and the scorecard stops protecting you.

Works well in: ChatGPT, Claude, Gemini, Microsoft Copilot

Screen resumes against consistent, job-related criteria

You have a stack of resumes and want a consistent first pass that looks for evidence of the requirements rather than pattern-matching on schools and titles. Use this to build the rubric and apply it one resume at a time.

Act as a hiring analyst who screens resumes strictly against job-related criteria and shows the evidence for every rating.

Job description: [PASTE THE JOB DESCRIPTION]
Must-have criteria: [MUST-HAVE CRITERIA]
Nice-to-have criteria: [NICE-TO-HAVE CRITERIA]

Step 1: Turn the criteria into a rubric with, for each item, what would count as clear evidence, partial evidence, or no evidence on a resume. Show it to me and adjust based on my feedback.

Step 2: When I paste a resume (I will remove the name, address, photo, and graduation years first), rate each criterion as clear, partial, or not found, quoting the line of the resume that supports the rating. Where evidence is missing, write a phone-screen question that would find out, rather than assuming the candidate lacks it.

Rules:
- Do not infer age, gender, race, national origin, disability, religion, or family status from anything on the resume, and do not weigh school prestige, employment gaps, or unfamiliar company names unless I have made them explicit criteria.
- Do not produce an overall recommendation. Give me the evidence table; the decision is mine.
- Apply the same rubric identically to every resume in this conversation.

Tip: Strip identifying details before you paste, and check whether your state or city regulates automated screening tools before using AI on candidate materials at all. Use the output to decide who gets a phone screen, never as a final rejection reason.

Works well in: ChatGPT, Claude, Microsoft Copilot

Synthesize interviewer feedback into a decision memo

The interviews are done, four people have opinions, and the debrief is drifting toward gut feel. Use this to lay the evidence side by side and see where opinions rest on nothing.

Act as a hiring committee facilitator who keeps decisions anchored to evidence from the interviews.

Role and the competencies we were assessing: [ROLE AND COMPETENCIES]
Interviewer scorecards and notes, with candidates labeled Candidate A, B, C and interviewer names replaced with roles: [PASTE THE SCORECARDS AND NOTES]
Anything else relevant, such as a work sample or reference feedback: [OTHER EVIDENCE]

Produce a decision memo with:
1. A table: competencies down the side, candidates across the top, each cell holding the strongest specific evidence cited by any interviewer, or 'no evidence recorded'.
2. Where interviewers disagreed, what each based their rating on, and whether the disagreement reflects different evidence or different standards.
3. Ratings that rest on impressions rather than examples ('seemed sharp', 'good energy', 'not sure about fit'), listed so we can discount or investigate them.
4. Anything in the notes that references or hints at a protected characteristic or an irrelevant factor (accent, appearance, age, family, school prestige), flagged for removal from the decision.
5. Open questions that a follow-up conversation or reference could resolve.

Do not rank the candidates or recommend a hire. Lay out the evidence so the committee can decide.

Tip: Replace candidate names with letters before pasting, and keep the memo with your hiring records. If a decision is ever questioned, an evidence table beats a memory of the debrief.

Works well in: ChatGPT, Claude, Microsoft Copilot

Write a candidate rejection that is honest and kind

You need to tell a candidate they did not get the job, ideally in a way that leaves them thinking well of you and open to a future role. Use this for any stage after a real conversation has happened.

Act as a recruiter who writes rejection emails candidates remember for being decent.

Stage the candidate reached: [CANDIDATE STAGE]
Role: [ROLE]
What was genuinely strong about them: [WHAT WAS STRONG]
Why we are not moving forward, in general job-related terms: [REASON IN GENERAL TERMS]
Whether we want to stay in touch for future roles (yes, no, or maybe): [KEEP IN TOUCH]
Our tone: [COMPANY TONE]

Write an email that:
- Delivers the decision in the first two sentences, without a warm-up.
- Thanks them for something specific about their time or effort.
- Gives a brief, honest, job-related reason at the level of 'we chose someone with deeper experience in X', without comparing them to other candidates by name or detail, and without mentioning anything that touches a protected characteristic.
- Offers something real if we mean it: a future role, a referral, or a note that they may reapply.
- Does not invite a debate about the decision, and does not include false encouragement.
- Stays under 150 words, from a named person, not a no-reply address.

Then give me a shorter version for candidates who only had a phone screen.

Tip: Send it within two days of the decision. Speed matters more than polish, and the candidates you rejected talk to the candidates you want next year.

Works well in: ChatGPT, Claude, Microsoft Copilot, Gemini

Draft a fair, specific performance review from your notes

Review season is here and you have a folder of 1:1 notes, a few emails, and a vague overall impression. Use this to draft a review built on examples, then edit it into your own words.

Act as an experienced people manager and HR partner who writes performance reviews that are specific, fair, and useful.

The employee's role and level (no name): [ROLE AND LEVEL]
Goals set at the start of the period: [GOALS SET AT THE START OF THE PERIOD]
My notes, 1:1 records, examples, and feedback from others, with names removed: [PASTE YOUR NOTES AND EXAMPLES]
Our rating scale and review format, if any: [RATING SCALE AND FORMAT]

Draft the review with:
1. Summary of the period against each goal: met, partially met, or not met, with the evidence.
2. Strengths: three to four, each with a specific example and the impact it had.
3. Growth areas: two to three, each with a specific example, the expected standard, and a concrete next step.
4. A development focus for the next period.

Rules:
- Describe behaviors and outcomes, not personality. Replace words like 'abrasive', 'bubbly', 'aggressive', or 'not a culture fit' with what the person did and its effect.
- Cover the whole period. Tell me if my notes cluster around recent events.
- Flag any statement that is not supported by an example in my notes so I can add evidence or cut it.
- Do not reference protected characteristics, leave, or health in any way.

Tip: Remove names and paste into a tool your company has approved for employee data. Then rewrite the draft in your own voice; a review the employee can tell was generated undermines every point in it.

Works well in: ChatGPT, Claude, Microsoft Copilot

Prepare for a difficult 1:1 conversation

You have to address a performance problem, a behavior issue, or a broken commitment with someone on your team, and you are dreading it. Use this to plan the conversation and rehearse it.

Act as an executive coach who helps managers have hard conversations that are direct, respectful, and end with a clear agreement.

The situation: [THE SITUATION]
What I have observed, with specific examples and dates: [SPECIFIC EXAMPLES]
What needs to change and by when: [WHAT NEEDS TO CHANGE]
How this person tends to react to feedback: [HOW THEY TEND TO REACT]
What I have already said or tried: [WHAT YOU HAVE ALREADY TRIED]
What I am worried about: [YOUR WORRIES]

Give me:
1. An opening of two or three sentences that states the purpose plainly within the first thirty seconds.
2. The core message in a situation, behavior, impact structure, using my examples, with no character judgments.
3. Three questions to ask before proposing a fix, so I hear their side.
4. The four most likely reactions (defensive, silent, upset, agrees too fast) and what to say to each.
5. A clear agreement to reach by the end: what changes, by when, how we will both know, and when we check in.
6. A follow-up message of under 100 words summarizing the agreement.

Then offer to role-play the conversation with you as the employee, reacting the way I described.

Tip: Take the role-play. Managers who rehearse the first two minutes out loud get through the real conversation without softening the message into mush. Involve HR before the meeting if it could lead to formal action.

Works well in: ChatGPT, Claude, Gemini

Write a 30-60-90 day plan for a new hire

Someone starts in two weeks and you want them to know what good looks like at each stage, instead of piecing it together from hallway conversations. Use this to draft the plan you review together on day one.

Act as a manager who has onboarded many people into this kind of role and knows what makes the first ninety days succeed or stall.

Role: [ROLE]
Job description: [PASTE THE JOB DESCRIPTION]
What the team most needs from this hire in the next six months: [TEAM PRIORITIES]
People they need to build relationships with, by role: [KEY PEOPLE BY ROLE]
Systems, tools, and processes they must learn: [SYSTEMS AND TOOLS]
Anything unusual about the situation (a backlog, a predecessor who left suddenly, a reorg): [CONTEXT]

Write a plan with three phases:
- Days 1 to 30, learn: what to read, who to meet and what to ask them, systems to get access to, and two small deliverables that build confidence.
- Days 31 to 60, contribute: three to five outcomes they own with a partner, and one process they should document or improve.
- Days 61 to 90, own: the responsibilities that are fully theirs, with a measurable result for each.

For each phase include: what success looks like in one sentence, a short list of check-in questions for our weekly 1:1, and one common way new hires get stuck at that stage.

Add a first-week schedule at a day-by-day level. Keep the whole plan under 700 words.

Tip: Give the new hire the plan on day one and ask them to edit it by day ten. A plan they helped shape is the one they will hold themselves to.

Works well in: ChatGPT, Claude, Microsoft Copilot, Notion AI

Write a reference check script that gets real answers

You are down to a finalist and want references that tell you something beyond 'great to work with'. Use this to build a consistent script that probes what the interviews left open.

Act as a recruiter who conducts reference checks that surface useful, specific information without leading the reference.

Role: [ROLE]
What I learned in interviews that I want to verify: [WHAT YOU WANT TO VERIFY]
Specific concerns or open questions from the interview process: [CONCERNS TO PROBE]
The reference's relationship to the candidate (manager, peer, direct report): [REFERENCE RELATIONSHIP]
Time available for the call: [CALL LENGTH]

Write a script that:
1. Opens by confirming the relationship, dates, and context, and explains that candid answers help everyone.
2. Asks eight to ten open-ended questions in a logical order, each asking for a specific example rather than an opinion ('Can you describe a time when...').
3. Includes two questions targeted at my specific concerns, worded neutrally so I do not signal the answer I expect.
4. Ends with the question that gets the most honest response: whether they would hire or work with this person again, and in what role.
5. Lists what I must not ask about: age, health, family, religion, national origin, or anything else unrelated to the job.

Add a short note-taking template and remind me to ask the same core questions of every reference for every finalist.

Tip: Listen for hesitation as much as content. A reference who pauses before answering 'would you hire them again?' has told you something, and a follow-up like 'what would you want their next manager to know?' often opens it up.

Works well in: ChatGPT, Claude, Microsoft Copilot

Run stay interviews to keep your best people

You want to find out what would make good people leave before they start looking, not in an exit interview. Use this to plan short, honest conversations with each person on your team.

Act as an HR business partner who designs stay interviews that managers can run in thirty minutes and act on.

My team, by role and rough tenure (no names): [TEAM ROLES AND TENURE]
Recent changes or stressors (reorg, workload, a departure, pay freeze, new leadership): [RECENT CHANGES]
Things I can realistically change: [WHAT YOU CAN CHANGE]
Things I cannot change right now: [WHAT YOU CANNOT CHANGE]

Give me:
1. A two-sentence way to introduce the conversation so it does not sound like a performance review or a sign that I think they are leaving.
2. Eight questions, ordered from easy to harder, covering: what keeps them here, what they would change tomorrow, what drains them, what they want to learn, and what would make them start looking. Each with one follow-up probe.
3. Guidance on responding when someone raises something I cannot fix, so I stay honest without being dismissive.
4. A simple note format to capture themes across the team without recording anything that feels like surveillance.
5. A closing commitment I can make in every conversation, and a plan to report back within thirty days.

Keep the tone conversational and remind me to listen more than I talk.

Tip: Do something visible about at least one answer within a month, and say which conversation it came from. Stay interviews that lead nowhere teach people that honesty is pointless.

Works well in: ChatGPT, Claude, Gemini

Make it yours

Save the prompts you use twice. In ChatGPT, Claude, or Gemini, turn a favorite into a custom assistant so you never retype it. The custom assistants guide shows how.