43% of organizations now use AI to support HR tasks, up from 26% the year before, according to SHRM's 2025 Talent Trends researchSource ↗
Recruiting is the top area for AI in HR (51% of organizations), and writing job descriptions is the single most common task at 66%, ahead of resume screening (44%) and candidate sourcing (32%)Source ↗
Nearly 9 in 10 HR professionals whose organization uses AI for recruiting (89%) say it saves time or increases efficiency, though 67% say their organization has not adequately trained people to work alongside AISource ↗
In SHRM's State of AI in HR 2026 survey of 1,722 HR professionals, recruiting led adoption at 27% while 54% of organizations reported no AI use in HR and no plans to adopt it in 2026Source ↗
writingChatGPTClaudeCopilot

Job descriptions that are inclusive and compliance-aware

Writing a job description from scratch is the most common HR AI task — 66% of recruiting teams do it — and the ones written under deadline tend to recycle inflated requirements, age-coded language, and boilerplate that quietly screens out strong candidates. AI turns a rough set of duties into a clean first draft in minutes. The real value is a prompt that also flags the legal and inclusivity problems a rushed human misses.

Prompt
You are an HR content assistant helping a hiring manager draft a job description. Draft a description for the role below that a qualified, diverse candidate pool would actually apply to.

Role: {{job_title}}
Team and reporting line: {{team_context}}
Must-have duties and requirements (use only these): {{core_requirements}}
Location, work model, and pay range: {{location_pay}}

Output in this order: a 2-3 sentence role summary, "What you'll do" (5-7 bullets), "What you'll need" split into Required and Preferred, and a one-line equal-opportunity statement placeholder.

Rules:
- Use only the duties and requirements I gave you. Do not invent responsibilities, certifications, degrees, or years of experience I did not list. Where something seems missing, add it to a separate "Questions for the hiring manager" list instead of guessing.
- Keep the Required list to genuine must-haves; move nice-to-haves to Preferred so we don't screen out non-traditional candidates.
- Flag any wording that could deter protected groups or raise ADA/EEOC issues (for example "young and energetic," "recent grad," unnecessary physical requirements) under a "Review for bias" note.
- Plain, inclusive language at a 9th-grade reading level. No jargon, no "rockstar/ninja," no inflated tone.
- Do not present the pay range as final if I have not confirmed it — mark it [CONFIRM].

Fill in your details and the prompt updates live — then copy.

planningClaudeChatGPT

Structured interview guides and scoring rubrics that reduce bias

Unstructured interviews are inconsistent and legally exposed: different candidates get different questions and gut-feel scores. HR can use AI to build a structured kit — the same job-related questions for every candidate plus an anchored 1-5 rubric — which improves hiring quality and creates the consistent, defensible record that EEOC guidance favors. It also keeps interviewers away from questions that touch protected characteristics.

Prompt
You are an HR assistant building a structured interview kit. Base everything on the job requirements I provide — do not assume duties I did not list.

Role: {{job_title}}
Key competencies to assess: {{competencies}}
Interview length and panel: {{format}}

Produce:
1. 6-8 behavioral and situational questions, each tied to one competency and phrased to draw out a specific past example (STAR-friendly). Job-related only.
2. For each question, 2-3 follow-up probes.
3. A scoring rubric: for each competency, a 1-5 scale with a concrete behavioral anchor describing what a 1, a 3, and a 5 answer look like.
4. An interviewer note listing question types to avoid (anything touching age, family or marital status, health or disability, religion, national origin, or arrest record) and why they are off-limits.

Rules:
- Every question must be job-related and legally safe. Do not generate questions about protected characteristics or that act as a proxy for them.
- Do not predict or assume how a candidate would answer.
- Keep it practical for a {{format}} interview.
- This kit supports human interviewers; final scoring and the hiring decision are made by people.

Fill in your details and the prompt updates live — then copy.

communicationChatGPTClaudeCopilot

Policy and benefits announcements employees actually understand

Every open enrollment, policy update, or benefits change means translating dense, legalistic source material into a clear message the whole company will read. It's high-volume writing that has to be accurate and on-tone. AI drafts the announcement fast — the discipline is keeping every factual claim tied to the official source document and getting the numbers checked before it goes out.

Prompt
You are an internal communications assistant for an HR team. Write a clear, warm company-wide announcement based only on the source details below. Do not add any benefit, date, or rule that isn't in the source.

Topic: {{topic}}
Key facts from the official source document: {{source_facts}}
Audience and tone: {{audience_tone}}
Action employees must take and the deadline: {{action_deadline}}

Format: a subject line, a one-line "the short version" at the top, a "what's changing and why" section, a "what you need to do" section with the deadline in bold, and a "where to get help" line.

Rules:
- Use only the facts I provided. If a detail employees would obviously ask about is missing, list it under "Gaps to fill before sending" rather than inventing it.
- Do not state any dollar amount, percentage, or date unless it is in my source facts — mark anything uncertain as [VERIFY].
- Plain English at an 8th-grade level, no legalese; define any benefits term you must use.
- Neutral and reassuring. Do not promise outcomes or give individual financial, tax, or legal advice.

Fill in your details and the prompt updates live — then copy.

analysisClaudeChatGPT

Turning open-text survey comments into themes leaders can act on

An engagement or pulse survey can return hundreds of free-text comments, and reading them into a coherent set of themes is slow, subjective work that often stalls before leadership ever sees the results. AI clusters the comments with representative quotes in minutes. The catch is that the comments are sensitive and frequently identifiable, so anonymization has to come first.

Prompt
You are a people-analytics assistant. I will paste anonymized open-text responses from an employee survey. Identify the main themes without exaggerating or inventing sentiment.

Survey question: {{survey_question}}
Anonymized responses: {{responses}}

Produce:
1. 5-8 themes, ranked by how often they appear. For each: a short label, an approximate share of comments, and 1-2 lightly paraphrased representative quotes.
2. Sentiment per theme (positive, mixed, or negative).
3. "Signal vs. noise": which themes are broad patterns versus one or two strong voices. Do not overweight a single vivid comment.
4. Suggested clarifying questions for a follow-up survey — not conclusions about individuals.

Rules:
- Base themes only on the responses provided. Do not infer who wrote a comment or attribute views to any person, team, or manager by name.
- If a response names a specific person, replace the name with [REDACTED] and flag it.
- Do not paraphrase in a way that changes meaning; preserve critical feedback honestly.
- Report counts as approximate ("about a third"), never as false precision.

Fill in your details and the prompt updates live — then copy.

writingClaudeChatGPT

Turning blunt manager notes into fair, specific review language

Managers hand HR rushed, vague, or harsh performance notes and expect polished, defensible write-ups. AI is good at reshaping raw notes into specific, behavior-based language. But this is exactly where people paste real names, ratings, and disciplinary details they shouldn't — and where the temptation to let AI "decide" the rating or the discipline is most dangerous.

Prompt
You are a writing coach for managers. Rewrite my rough performance-feedback notes into clear, specific, behavior-based language for a review. Keep my assessment; improve the wording only.

Role level: {{role_level}}
My rough notes (behaviors and examples, names replaced with placeholders): {{rough_notes}}
Review period and the rating I have already decided: {{period_rating}}

For each point, produce: the behavior (observable and specific), the impact, and a forward-looking suggestion. Use "situation - behavior - impact" framing.

Rules:
- Do not change my rating or invent accomplishments, incidents, or metrics I didn't mention. If a claim is vague ("bad attitude"), rewrite it as a request for a specific example rather than stating it as fact.
- Keep it professional and non-judgmental about the person; focus on work behavior, not personality or protected characteristics.
- Remove language that could read as discriminatory or retaliatory, and flag what you removed.
- Do not recommend or decide any disciplinary action, promotion, or termination — that is my decision with HR. Balance developmental and positive feedback.

Fill in your details and the prompt updates live — then copy.

automationChatGPTClaudeCopilot

First-draft answers to repetitive policy questions from your handbook

HR fields the same policy questions over and over — PTO accrual, remote-work rules, expense limits — and typing individual answers is repetitive drudgery. Feeding AI your actual handbook lets it draft consistent, on-policy answers you can reuse in an FAQ or knowledge base. The rule is strict: the AI must answer only from your document, never from its own idea of "standard" policy.

Prompt
You are drafting internal HR FAQ answers. Answer each employee question using ONLY the policy text I provide. Do not use general knowledge of typical company policies.

Policy source text: {{policy_text}}
Questions to answer: {{questions}}
Company voice: {{voice}}

For each question: a short, plain-English answer, then a "Policy reference" line citing the section of my source it came from.

Rules:
- Answer strictly from the provided policy text. If the answer isn't in it, write "Not addressed in the current policy — routing to HR" instead of guessing or filling from general practice.
- Do not give legal, tax, or individual benefits advice. For anything involving accommodations, leave (FMLA/ADA), pay disputes, or terminations, direct the employee to a named HR contact.
- Flag any place where the policy is ambiguous or seems to contradict itself under "For HR/legal review."
- Keep each answer under 120 words and neutral in tone.

Fill in your details and the prompt updates live — then copy.

Common questions from hr managers

Is it okay for HR to use ChatGPT or Claude on employee-related work?

Yes, with guardrails — adoption is mainstream, and SHRM found 43% of organizations already use AI in HR. The conditions are consistent: no confidential employee data in consumer accounts that train on inputs, a human decision on anything that affects someone's job, and legal review of policies and formal notices. Check your organization's AI policy before your first prompt, not after.

Can I paste employee names, performance ratings, or medical information into an AI tool?

Not into consumer accounts. They retain and may train on what you paste, and employee records, health data, and disciplinary details are confidential and often regulated. Use placeholders and generic descriptions, or an approved enterprise tool with a no-training and retention agreement plus your organization's sign-off. When in doubt, keep it in your HRIS.

Can AI make hiring, discipline, or termination decisions to save time?

No. Anti-discrimination law holds the employer liable for biased outcomes even when a vendor's AI produced them, and automated rejection tools face bias-audit laws and lawsuits such as Mobley v. Workday. AI can draft, organize, and summarize, but a human must make and document the decision — with legal review for terminations and investigations.

Can I use AI to help run a workplace investigation?

Only for support tasks, and carefully. AI can help you structure an interview outline or tidy your own notes, but it cannot judge credibility, decide outcomes, or replace a legally sound process — and investigation records are highly confidential. Keep names and allegations out of consumer tools, and have counsel review any conclusions before you act on them.

Related professions