Best AI Prompts for HR

AI summary

Seven AI prompts for HR business partners, HR generalists, and people operations leads: policy update drafting, PIP frameworks, investigation memos, termination conversation prep, benefits open enrollment comms, new-hire onboarding plans, and compensation conversation prep. Built to scaffold judgment without crossing into legal advice. Every output is HR-reviewed; AI never produces legally binding language alone.

Prompt 1

Policy Update Drafter

Most policy updates get written under deadline and circulated without proper review. This prompt drafts the update so legal review focuses on substance, not form.

I need to draft a policy update on:

TOPIC: [SPECIFIC: remote work, harassment, PTO, AI use, etc.]
WHY THIS UPDATE NOW: [TRIGGER]
WHAT IS CHANGING from prior policy: [SPECIFIC]
WHO IT AFFECTS: [SCOPE]
LEGAL CONSTRAINTS I am aware of (state law, federal, NLRA, ADA, etc.): [LIST]
LEADERSHIP'S STATED INTENT for the update: [BRIEF]

Draft the policy section:

1. PURPOSE: one sentence on why this policy exists.
2. SCOPE: who it covers, who it does not.
3. THE POLICY ITSELF: in plain English, with specific actions employees should and should not take.
4. THE PROCESS: how to raise questions, request exceptions, or report violations.
5. CONSEQUENCES: stated factually, calibrated to severity.
6. EFFECTIVE DATE and transition guidance.
7. THE 3-5 QUESTIONS this policy will trigger from managers and employees, with proposed answers.

Do not invent legal references. Do not name specific statutes or regulations; I will add citations with employment counsel. Tone: clear, non-punitive, never legalistic where plain language works.

When to use: When leadership requests a policy change or compliance trigger appears. · Best model: Claude. Discipline about not naming specific statutes matters.

Prompt 2

Performance Improvement Plan Framework

PIPs handled badly create legal exposure. This prompt drafts the framework so the conversation, the document, and the follow-through align.

Employee situation (de-identified):

ROLE / LEVEL: [BRIEF]
LENGTH OF TENURE: [DURATION]
MANAGER'S DOCUMENTED PERFORMANCE CONCERNS: [SPECIFIC, OBSERVABLE]
PRIOR CONVERSATIONS about performance: [WHEN, WHAT WAS SAID, WHAT WAS WRITTEN]
WHAT THE EMPLOYEE HAS BEEN TOLD they need to improve: [BRIEF]
WHAT THE EMPLOYEE HAS NOT BEEN CLEARLY TOLD that we now want to formalize: [BRIEF]
THE SUCCESS CRITERIA we want to set: [SPECIFIC, MEASURABLE]

Draft a PIP framework:

1. THE CONVERSATION OPENING: how the manager introduces the PIP without blindsiding the employee.
2. THE PERFORMANCE GAPS: stated factually with specific observable evidence.
3. THE EXPECTATIONS: behavioral, measurable, time-bound. Not vague.
4. THE SUPPORT: what the company provides (training, mentorship, regular check-ins).
5. THE CHECK-IN CADENCE: weekly or biweekly, with what each check-in covers.
6. THE CONSEQUENCES: what happens if improvement is achieved, partial, or absent.
7. THE DOCUMENT STRUCTURE: section-by-section.

FLAG anything in my inputs that suggests the PIP should not happen yet (insufficient documentation, missing feedback history, possible protected-category retaliation concern).

This is a draft for HR and legal review. Do not skip the FLAG step.

When to use: After manager and HR have agreed a PIP is warranted, before drafting the document. · Best model: Claude. The discipline about flagging premature PIPs matters legally.

Prompt 3

Employee Investigation Memo Drafter

Investigation memos must be factual, contemporaneous, and defensible. This prompt structures the memo from your notes so the documentation holds up later.

Investigation summary (de-identified):

DATE OF COMPLAINT: [DATE]
NATURE OF COMPLAINT: [BRIEF, FACTUAL]
INVESTIGATOR: [ROLE]
WITNESSES INTERVIEWED: [ROLES, NOT NAMES]
DOCUMENTS REVIEWED: [LIST OF CATEGORIES]
FINDINGS: [WHAT WAS CONCLUDED]
LIMITATIONS of the investigation: [BRIEF]
NEXT STEPS taken or recommended: [BRIEF]

Draft an investigation memo:

1. PURPOSE: one sentence on what the investigation addressed.
2. SCOPE: what was investigated, what was not.
3. METHODOLOGY: how the investigation was conducted, including any limits.
4. FINDINGS: factual, in the order they were established.
5. EVIDENCE: what supports each finding (interview type, document category; do not quote names).
6. CONCLUSIONS: drawn from the findings, stated with appropriate certainty.
7. RECOMMENDATIONS: actions HR or leadership should consider.
8. RETENTION: how long this record should be retained and where it is filed.

Do not invent witnesses, documents, or findings. Preserve uncertainty where it exists ("the complainant reported," "the witness recalled," "the email shows"). This document may be subpoenaed; write accordingly.

When to use: Within 5 business days of the investigation closing. · Best model: Claude. The discipline about preserving uncertainty matters legally.

Prompt 4

Termination Conversation Prep

Terminations done badly create legal and human harm. This prompt prepares the conversation so the words are right and the process holds.

Termination decision (de-identified):

EMPLOYEE TENURE: [DURATION]
CATEGORY of termination: [PERFORMANCE / CONDUCT / LAYOFF / VOLUNTARY-AGREED]
DOCUMENTATION supporting the decision: [BRIEF]
LEGAL REVIEW COMPLETED: [YES / IN PROGRESS]
SEVERANCE / BENEFITS: [WHAT IS BEING OFFERED]
NOTIFICATIONS to others (manager, team, payroll, IT, security): [WHAT IS PLANNED]
KNOWN SENSITIVITIES (recent protected-status disclosure, recent complaints filed, recent FMLA, etc.): [FLAG IF ANY]

Draft a termination conversation framework:

1. THE OPENING: factual, brief, does not start with apology or pleasantries.
2. THE DECISION stated cleanly in one sentence.
3. THE REASON stated in a way that aligns with the documentation (do not invent reasons not in the documentation).
4. THE LOGISTICS: effective date, last day, severance overview, benefits, references.
5. THE NEXT 30 MINUTES: what happens immediately after (IT, security, return of property).
6. THE EMPLOYEE'S OPTIONS: severance signing window, COBRA, unemployment, references policy.
7. THE CLOSE: dignified, without softening that suggests negotiability.

FLAG IMMEDIATELY: if any of the known sensitivities trigger heightened legal risk (recent FMLA, recent complaint, recent disclosure), do NOT proceed without explicit employment counsel sign-off. State this in the output, regardless of how the inputs framed the situation.

This is the framework. The conversation will be conducted by an HR partner trained in termination procedure.

When to use: 24-48 hours before the conversation, after legal and HR alignment. · Best model: Claude. The discipline about flagging heightened legal risk is essential.

Prompt 5

Benefits Open Enrollment Communication

Most benefits communication during open enrollment goes ignored. This prompt drafts the message so employees actually act.

Open enrollment period:

WHAT IS NEW this year: [SPECIFIC CHANGES TO PLANS, CONTRIBUTIONS, NETWORK]
KEY DEADLINES: [DATES]
WHAT EMPLOYEES MUST DO actively: [SPECIFIC]
WHAT WILL HAPPEN if they take no action: [DEFAULT BEHAVIOR]
KEY DECISIONS that affect employees most: [LIST]
NEW BENEFIT or option worth highlighting: [BRIEF]

Draft an open enrollment email series:

1. ANNOUNCEMENT EMAIL (3 weeks before deadline): what is changing, why it matters, when to act.
2. EDUCATIONAL EMAIL (2 weeks before): walk through the key decision (HDHP vs PPO, HSA contribution, life insurance level).
3. REMINDER EMAIL (1 week before): clear deadline, what happens if no action, where to get help.
4. FINAL CALL EMAIL (2 days before): direct, brief, includes default behavior warning.
5. CONFIRMATION EMAIL (after deadline): what their elections are, when they take effect, how to fix mistakes.

Keep each email under 250 words. Use plain English; assume readers are not benefits experts. Include the one specific number that matters most for each decision (deductible, contribution match, premium).

Do not invent benefit details. Do not use "valuable" or "comprehensive" without specificity.

When to use: 8 weeks before open enrollment opens. · Best model: Claude or ChatGPT. Both handle this structural work.

Prompt 6

New-Hire Onboarding Plan

Most new-hire experiences are inconsistent. This prompt drafts a 30/60/90 plan tailored to the role rather than a generic template.

New hire (de-identified):

ROLE: [TITLE, LEVEL]
TEAM: [BRIEF]
MANAGER: [ROLE]
EXPECTED FIRST-90-DAY CONTRIBUTION: [WHAT SUCCESS LOOKS LIKE]
KEY SYSTEMS they will need: [LIST]
KEY PEOPLE they should meet: [ROLES]
DOMAIN KNOWLEDGE they need to build: [LIST]
ORG CULTURE notes they should learn: [BRIEF]

Draft a 30/60/90 onboarding plan:

1. WEEK 1: setup, orientations, first meetings. Avoid the trap of all-day shadowing.
2. WEEKS 2-4: role-specific ramp activities tied to expected contribution.
3. DAYS 30-60: first deliverables, calibration with manager, expanding network.
4. DAYS 60-90: leading something small, contributing to team rituals, calibrating against expectations.
5. CHECKPOINTS: explicit 30/60/90 conversations with what each covers.
6. MENTORSHIP: a recommended buddy pairing if role/seniority suggests it.
7. THE "YOU ARE BEHIND" SIGNALS: what would tell us at day 30 we need to course-correct.
8. THE "YOU ARE AHEAD" SIGNALS: what would tell us this hire is exceeding expectations.

Do not invent processes our company does not have. Build only from what I described.

When to use: Two weeks before the new hire starts. · Best model: Claude. The discipline about role-specific tailoring matters.

Prompt 7

Compensation Conversation Prep

Comp conversations go badly when HR walks in unprepared. This prompt structures the prep so the conversation is grounded, fair, and consistent.

Compensation conversation context (de-identified):

EMPLOYEE TENURE and recent performance: [BRIEF]
CURRENT COMP: [SALARY BAND, BONUS, EQUITY]
WHAT THEY ARE ASKING FOR or what we are offering: [SPECIFIC]
COMP BAND for the role: [RANGE]
INTERNAL EQUITY considerations: [HOW THIS COMPARES TO PEERS]
MARKET DATA we are using: [SOURCE]
WHAT WE CAN MOVE ON (base, bonus, equity, title): [WHAT IS FLEXIBLE]
WHAT WE CANNOT MOVE ON: [HARD CONSTRAINTS]

Draft the conversation prep:

1. THE OPENING: warm, grounded, acknowledges the conversation matters.
2. THE FACTS WE WILL SHARE: market data, band, internal equity, recent performance acknowledgment.
3. THE NUMBER OR OFFER stated cleanly when we get to it.
4. THE REASONING: what the offer is grounded in, in their language.
5. THE LIKELY OBJECTION or counter, and how to respond.
6. THE TRADE-OFFS available: where we can move if they push.
7. THE NEXT STEPS: timing, what they should do, when we will follow up.

FLAG any inputs that suggest pay equity concerns (the request is significantly outside the band relative to peers with similar tenure and performance) and recommend HR partner consult before the conversation.

Tone: warm, direct, never paternalistic. Comp conversations are about respect.

When to use: Before the comp meeting. · Best model: Claude. Tone discipline matters.

Connectors are now standard

Claude, ChatGPT, and Grok all support connectors that let your AI read live data from your work tools (Gmail, Notion, GitHub, Asana, HubSpot, Stripe, and many more) instead of relying on you to paste context. For HR this means the AI can read your Notion or SharePoint HR policy library, your Workable / Ashby / Greenhouse pipeline data, your Gmail or Outlook thread with managers, or your Calendly meeting bookings. Do not connect AI directly to HRIS systems holding PII without a vetted enterprise agreement.

For HR professionals, the connectors worth pairing with these prompts:

Get Smarter About AI Every Morning

Free daily newsletter. Built for people who want to use AI well, not chase every model.

Free forever. Unsubscribe anytime.

Two ways to go further