Skip to content
Playbook
IntermediatePillarHRPeopleOpsGeneralist

Prompt libraries for HRBPs: the 40 prompts you'll actually use

A practical prompt library for HR business partners — vetted prompts for hiring, performance, comp, employee relations, policy drafting, and analytics.

24 min read
On this page
60-Second Summary
  • A good HR prompt = role + task + context + format + constraints + verification step. Five sentences. Every prompt below follows this pattern.
  • The biggest unlock isn't a clever prompt — it's a saved library of prompts your team uses repeatedly. Standardization beats cleverness.
  • Three things to NEVER paste into a public LLM: real employee names + sensitive info, comp tables, anything covered by GDPR/DPDP/HIPAA. Use enterprise tools (ChatGPT Enterprise, Claude for Work, Copilot with retention disabled) or anonymize.
  • AI is a thought partner, not an authority. Every output needs a human pass for accuracy, tone, and legal exposure — especially employee relations and policy drafts.

Most HR teams using AI use it badly — typing one-line questions into ChatGPT and getting generic answers. The teams that get real leverage have a shared prompt library: 30–60 reusable prompts that the whole HR team uses. This guide shows the design principles, then gives you 40 starter prompts across every HR domain.

Prompt design rules

The six elements of a useful HR prompt
  1. 1
    1. Role
    Tell the model who it's pretending to be. 'You are an experienced HR business partner at a 200-person tech company.'
  2. 2
    2. Task
    State the specific task in one sentence. 'Draft a performance improvement plan for the situation described below.'
  3. 3
    3. Context
    Provide the data the model needs. Anonymize names. Include enough specifics that the output isn't generic.
  4. 4
    4. Format
    Specify output structure. 'Return as: (a) summary, (b) 3 measurable goals, (c) 30/60/90 milestones, (d) suggested manager script.'
  5. 5
    5. Constraints
    Set guardrails. 'Keep under 400 words. Avoid legal terminology that requires lawyer review. Note any assumptions you're making.'
  6. 6
    6. Verification
    Ask the model to flag what to double-check. 'At the end, list any factual claims I should verify before sending.'
What NOT to paste into AI

Public LLMs (free ChatGPT, free Claude, free Gemini) may train on your input. Never paste: real employee names + sensitive content (performance issues, health, harassment), full comp tables, internal-only strategy docs, PII covered by GDPR/DPDP/HIPAA, board materials, candidate data without consent. Use enterprise tools (ChatGPT Enterprise, Claude for Work, Microsoft 365 Copilot with M365 commercial data protection) for sensitive work — they contractually exclude training. When in doubt, anonymize.

Hiring prompts

1. Draft a job description from a role brief

Use case: turn a 5-line role brief from a hiring manager into a complete, bias-checked JD in 3 minutes.

Prompt

You are a senior recruiter writing job descriptions for a mid-sized tech company. Draft a JD for the role described below. Output: (1) role title, (2) one-paragraph company line (leave [Company description] placeholder), (3) what you'll do (5 bullets), (4) what we're looking for (5 must-haves, 3 nice-to-haves), (5) what we offer (placeholder), (6) inclusive closing. Constraints: avoid masculine-coded language (Gaucher & Friesen lexicon), avoid 'rockstar/ninja/guru', avoid years-of-experience hard floors except where legally needed. Flag any assumption you made. Role brief: [paste brief].

2. Interview question generator

Prompt

You are an interview-design specialist. Generate 8 behavioral interview questions for the role and competency below. Use the STAR structure. Each question should test one competency, be open-ended, and avoid hypotheticals. Include a 3-tier rubric (excellent/expected/concerning) with concrete signal markers. Competency: [e.g., dealing with ambiguity]. Role: [title + level].

3. Resume summarization for hiring managers

Prompt

You are an HRBP preparing a hiring manager for a screen. Summarize the resume below in: (a) 3-sentence narrative, (b) top 3 strengths against the JD, (c) top 3 gaps or questions to probe, (d) one career-trajectory observation. Do not infer demographic information. JD: [paste]. Resume: [paste, anonymized].

4. Candidate rejection email (warm)

Prompt

You are an HR coordinator writing a rejection email after a final-round interview. Tone: warm, specific, brief (under 150 words). Mention one genuine positive from their candidacy. Don't promise future contact unless specified. Don't give specific feedback that could be litigated. Candidate role: [title]. Stage rejected at: [stage]. One positive to mention: [bullet].

5. Sourcing search string

Prompt

You are a recruiter writing Boolean search strings for LinkedIn Recruiter. Generate 3 search-string variants for the role below: (1) tight match, (2) adjacent-skill expansion, (3) underutilized synonyms. Explain the difference in expected pool size. Role: [title, location, must-have skills, exclusions].

Performance prompts

6. Convert manager notes into a review draft

Prompt

You are an HRBP coaching a manager on writing performance reviews. Below are the manager's rough notes about an employee for a half-year review. Convert into a structured review draft with: (a) headline assessment (2 sentences), (b) what went well (3 specific examples), (c) areas to grow (2 specific examples, future-focused), (d) goals for next half (3 SMART goals). Use observable behavior language; avoid character judgments. Notes: [paste, employee name = 'Employee'].

7. Feedback rephraser (sharper, kinder, or clearer)

Prompt

You are a feedback coach. Take the feedback below and rewrite it three ways: (1) more direct (cuts hedging), (2) more empathetic (acknowledges effort first), (3) more specific (replaces vague claims with observable behavior). Explain when each version is appropriate. Original feedback: [paste].

8. PIP draft

Prompt

You are an HRBP at a 150-person tech company. Draft a 60-day Performance Improvement Plan based on the situation below. Include: (1) one-paragraph context, (2) three SMART improvement goals with measurable success criteria, (3) 30/60-day check-in milestones, (4) support resources from the manager, (5) what 'not meeting' the PIP looks like (consequences). Flag where employment-law counsel review is recommended. Situation: [anonymized notes].

9. Calibration meeting prep

Prompt

You are an HRBP facilitating a calibration meeting for engineering. Below are 6 employees' draft ratings and 2-sentence justifications. Identify: (a) likely calibration debates (gaps between draft ratings and stated evidence), (b) suspected leniency or strictness patterns by manager, (c) demographic representation concerns at top and bottom of the distribution, (d) suggested discussion order. Draft ratings: [anonymized table].

Comp prompts

10. Comp band rationalization

Prompt

You are a compensation analyst. Compare our internal band for [role] (P25/P50/P75 = $X/$Y/$Z) with the public benchmarks below. Identify: (a) overall position vs market (lag/match/lead), (b) tightest band (where compression risk is highest), (c) recommended adjustments. Show your math. Internal: [paste]. Benchmarks: [paste 3 sources].

11. Offer-letter narrative

Prompt

You are a recruiter writing the personal-context paragraph that accompanies an offer letter. Below are: (a) candidate's stated motivations from the final interview, (b) the offer numbers. Write a 100-word paragraph framing the offer in terms of their motivations. No new factual claims; only reframing of stated offer terms. Inputs: [paste].

12. Total-rewards comparison for a candidate question

Prompt

You are a comp & benefits specialist. A candidate asks how our offer compares to their current employer. Below are: (a) our offer, (b) their current package. Produce a structured comparison covering base, bonus, equity (with assumptions noted), benefits, leave, total cash year 1, total compensation year 4. Note where comparison is meaningful and where it is apples-to-oranges. Inputs: [paste].

13. Pay-equity audit briefing

Prompt

You are a senior comp analyst summarizing a pay-equity audit for the CEO. The audit (run by [tool]) found the following: [paste anonymized headline findings]. Produce: (1) one-page executive summary, (2) top 3 risks, (3) recommended actions with rough cost, (4) communications guidance. Avoid speculative legal claims; recommend counsel review where appropriate.

Employee relations prompts

14. Manager script for a difficult conversation

Prompt

You are a senior HRBP coaching a manager on a difficult conversation. The situation: [anonymized summary]. Produce: (1) opening script (3 sentences, sets context without ambushing), (2) the core message (1 paragraph), (3) anticipated employee responses with manager replies (3 scenarios), (4) close script. Tone: direct + respectful. Flag any topic that requires HR or legal presence in the room.

15. Investigation timeline drafter

Prompt

You are an ER specialist. Given the anonymized complaint below, draft an investigation plan: (1) immediate actions (24h), (2) witnesses to interview with rationale, (3) documents to preserve, (4) timeline (target close in 14 business days), (5) interim measures to consider. Note jurisdiction-specific procedural requirements to confirm with counsel. Complaint: [paste].

16. Documentation cleanup

Prompt

You are an ER lead reviewing case notes. Convert the manager's narrative below into structured ER documentation: (a) date + parties + role context, (b) factual observations (separated from inferences), (c) statements made (verbatim where possible), (d) actions taken, (e) follow-up scheduled, (f) flagged inferences or assumptions to verify. Notes: [paste].

Policy & communication prompts

17. Policy first draft

Prompt

You are a senior People Ops lead drafting policy for a 150-person company operating in [countries]. Draft a [policy name, e.g. 'Remote Work Policy']. Include: (1) purpose, (2) scope, (3) eligibility, (4) responsibilities, (5) procedure, (6) exceptions, (7) approvals, (8) review cadence. Keep under 700 words. Use plain language. Flag jurisdiction-specific items that require local counsel review.

18. All-hands announcement

Prompt

You are a communications partner to the CEO. Below are bullet points for an internal announcement. Write a 250-word all-hands message that: (a) leads with what changes for the employee, (b) gives the reasoning, (c) acknowledges hard parts honestly, (d) closes with a clear next action. Tone: direct, human, neither corporate nor casual. Bullets: [paste].

19. FAQ generator

Prompt

You are an HR comms partner. Below is a new policy or announcement. Generate the 12 most likely employee questions, grouped by theme, with draft answers in plain language. Flag any answer that requires a leader or lawyer to finalize before publishing. Source: [paste].

20. Handbook language modernizer

Prompt

You are a People Ops writer updating a handbook section. Rewrite the language below to be: (a) plain English (Grade 9 reading level), (b) inclusive (gender-neutral, accessible), (c) action-oriented (what the employee does, not the company asserts). Preserve all legal content. Section: [paste].

Analytics & reporting prompts

21. Attrition narrative from numbers

Prompt

You are a people analytics lead presenting to the leadership team. Below are quarterly attrition numbers by team. Produce: (1) 3-sentence executive summary, (2) top 3 patterns worth attention, (3) likely vs. confirmed causes (label each), (4) recommended next questions to investigate. Numbers: [anonymized table].

22. Engagement survey analysis

Prompt

You are a senior People Ops analyst. Below are open-text responses (anonymized) from an engagement survey question. Categorize into 5–7 themes with frequency counts. For each theme: (1) summary, (2) representative quote, (3) suggested action category (process, manager, comp, strategy, culture). Responses: [paste].

23. Diversity dashboard explainer

Prompt

You are a DEI analyst. Below are representation numbers at each level. Produce a narrative for a leadership review: (a) where representation matches/exceeds parity, (b) where it lags, (c) likely root cause (hiring funnel, attrition, promotion) and what data we'd need to confirm, (d) two interventions with evidence base. Numbers: [paste anonymized].

Building your team's library

  1. Start with 10 prompts the team uses weekly. Save in a shared Notion / Confluence / docs page.
  2. Tag prompts by domain (Hiring / Performance / Comp / ER / Policy / Analytics) and outcome (Draft / Analyze / Communicate / Decide).
  3. Each prompt has: title, when to use, the prompt itself, what to verify after, example before/after.
  4. Run a 30-minute monthly 'prompt session' — every HR team member shares one new prompt and one prompt that didn't work.
  5. Maintain a 'do not paste' list of types of content that must be anonymized first.
  6. Audit quarterly: which prompts are used? Which are stale? Which produced quality issues?
  7. Build prompts for your tools (HRIS exports, ATS reports, engagement survey formats) — context that's repeatable.

FAQ

Frequently asked questions

Which AI tool should HR use?

For sensitive work: ChatGPT Enterprise, Claude for Work, or Microsoft 365 Copilot — all contractually exclude training on your data. For non-sensitive work, paid Claude/ChatGPT are fine. Free tiers are not appropriate for any work involving employee information.

Should we ban personal ChatGPT use for HR work?

Don't ban — guide. A policy that allows AI for non-sensitive tasks (drafting public job descriptions, brainstorming) and prohibits it for sensitive work (real employee data, comp, ER) is enforceable and realistic. An outright ban is widely ignored and harder to enforce than a clear guide.

How do we know if an AI output is wrong?

Three signals: (1) confident assertions with no source, (2) statistics that round suspiciously well (always 'a 2024 study found 73%'), (3) legal claims without jurisdiction. Treat every output as a first draft. Verify before sending.

Do these prompts work in any language?

Mostly yes, with caveats. The major models (GPT-4 class, Claude 3 class, Gemini 1.5+) handle English, Spanish, French, German, Mandarin, Hindi, Portuguese well. They handle Nepali, Bengali, and Tagalog adequately. Always verify outputs in less-common languages — hallucination rates are higher.

How do we measure if our prompt library is working?

Three metrics: (a) % of HR team using the library weekly (target: 80%+), (b) time saved per use case (track 3 use cases for 4 weeks before/after), (c) quality issues raised (target: <5% of outputs need substantial rework).

Written by Pawan Joshi.Sources cited inline.
First published 15 Jun 2026See site changelog →