Most of HR's writing is first drafts: a handbook section for a new state, the open-enrollment announcement, manager talking points for a reorg, a job description copied forward since 2019. A chat assistant produces a competent first version in under a minute, so your hour goes to editing.
Second, analysis you could never afford. Two thousand open-text engagement comments used to get skimmed and summarized from memory. Paste them, anonymized, into a tool with a large context window and you get themes, counts by department, and checkable quotes.
What AI does not change: you own the decision. Hiring, discipline, accommodation, and termination carry legal exposure, and laws like New York City's Local Law 144, Illinois' Human Rights Act amendments, and Colorado's AI Act treat AI in employment decisions as regulated activity. Use AI to prepare and draft, never to decide.
Quick wins this week
- Paste your PTO policy and ask for a plain-English rewrite at an eighth-grade reading level. Compare it to the original clause by clause before publishing.
- Turn a rambling manager complaint into a neutral, dated summary of facts and open questions before deciding whether it needs a formal investigation. Swap names for Employee A and B first.
- Ask for a job description rewrite that removes age-coded and gendered language ('digital native', 'rockstar', 'recent graduate') and keeps only requirements you would actually screen on.
- Feed in last quarter's anonymized exit-interview notes and ask for the five biggest themes with one supporting quote each. Confirm each quote exists in the source.
What AI can do for hr professionals, task by task
Policy and handbook drafting
Give the assistant the current text, the change you need, your states, and the tone. Ask for the new version plus a list of every change and why. Check each legal point against counsel or a dated law-firm summary; models get state specifics wrong.
Employee communications and change announcements
Paste the facts (what changes, when, who is affected, what does not) and name the audience. Ask for the announcement, manager talking points, and an FAQ in one pass. Cut anything that sounds like a promise you cannot keep.
Engagement and exit survey analysis
Strip names from the open-text export and paste it with the question and department labels. Ask for themes with approximate counts, three verbatims each, and what is new since last cycle. Check that every verbatim exists word for word; models round counts and invent quotes.
Investigation timelines and note organization
Paste your own typed interview notes with names replaced by roles and ask for a dated timeline, the points where accounts conflict, and open questions. You write the findings yourself. If counsel directs the investigation, ask before using any AI tool.
Compliance research and state comparisons
Ask for a table of one requirement across your states, such as pay-range disclosure in postings, with statute, headcount threshold, and effective date per row. Treat it as a list of things to verify, and confirm every row against the statute.
HR knowledge base and self-service answers
Load your handbook and benefits guides into a tool that answers only from your documents: NotebookLM, a Claude Project, a custom GPT, or Copilot on SharePoint. Test it with the twenty questions your inbox gets most. Leave, accommodation, and complaint questions always route to a person.
Prompts for hr professionals
Replace the bracketed placeholders, paste into any chat assistant, and iterate on the result.
Rewrite a policy in plain English
You are an HR policy writer for busy employees, not lawyers. Rewrite the policy below so an eighth-grader could follow it, without changing any rule, number, or eligibility condition. Policy: [PASTE THE POLICY TEXT] Context: [COMPANY SIZE, STATES, UNION OR NON-UNION] Return the rewritten policy with the same headings, then a table of every place you changed meaning or were unsure. Flag ambiguous sentences instead of guessing. Add no legal citations I did not include.
Tip: Send the flagged table to counsel; it is often the most useful part.
Theme open-text survey comments
You are an HR analyst. Below are anonymized responses to '[PASTE THE SURVEY QUESTION]', each tagged with a department. Identify the 5 to 8 most common themes. For each, give a label, an approximate count, where it concentrates, three verbatim quotes, and the tone. Then list rare comments a leader should still see (safety, harassment, legal concerns). Never paraphrase or invent quotes. Responses: [PASTE THE COMMENTS]
Tip: Batch large comment sets, then ask for a merged theme list.
Draft a change announcement with manager talking points
You are an internal communications specialist supporting HR. Draft (1) an all-employee announcement under 250 words, (2) manager talking points covering the five hardest questions, and (3) an 8-question employee FAQ. What is changing: [DESCRIBE THE CHANGE] Effective date: [DATE] Who is affected: [GROUPS OR LOCATIONS] What is not changing: [LIST] Tone: [WARM AND DIRECT / FORMAL / BRIEF] Promise nothing beyond these facts; where one is missing, write [NEEDS CONFIRMATION] instead of guessing.
Tip: Search for 'NEEDS CONFIRMATION' before anything goes out.
Plan a difficult conversation and role-play it
You are an experienced HR business partner coaching a manager. Situation: [DESCRIBE THE SITUATION, HISTORY, AND GOAL]. Manager's experience: [FIRST-TIME / EXPERIENCED]. Known risk factors: [RECENT COMPLAINTS, PROTECTED LEAVE, ACCOMMODATION REQUESTS, OR NONE]. Give a plan: purpose in one sentence, an opening line, three key messages, likely reactions and responses, phrases to avoid, and next steps. Then role-play the employee while I practice, and afterward critique my tone, clarity, and legal risk.
Tip: Honest risk factors are what make the legal-risk critique useful.
State compliance comparison table
You are an HR compliance researcher. Build a comparison table of [THE REQUIREMENT, E.G. PAY RANGE DISCLOSURE] for these states: [LIST OF STATES]. Columns: state, whether it applies, employer-size threshold, what must be disclosed, penalties, statute name, effective date, your confidence (high, medium, low), and the exact source I should read. Write 'unknown' where unsure, and end with recent changes I should check since your knowledge may be stale.
Tip: Every cell, even 'high confidence', stays unverified until you read the source.
Want a prompt for something else? Use the Prompt Builder or browse education and training prompts, hiring and management prompts, leadership and communication prompts.
Skills to build
Writing a full brief before you ask
Why: Generic inputs produce generic text you rewrite; the assistant needs the audience, states, history, and what must not change.
How: Use one template every time: role for the AI, audience, facts, constraints, format, and 'ask me questions first if anything is unclear'.
Anonymizing by reflex
Why: Names, IDs, medical details, and pay figures must never reach a consumer AI tool, and a habit beats a rule remembered under deadline.
How: Replace names with roles and delete IDs, birth dates, and medical language before every paste. Keep a red list of data you never paste.
Verifying legal claims against a dated source
Why: Models cite repealed statutes, wrong thresholds, and stale effective dates with total confidence.
How: For every legal statement in AI output, find a primary or law-firm source dated within the past year, and log the mistakes you catch.
Knowing AI-hiring law well enough to question vendors
Why: Your ATS, assessment, and video-interview vendors already use AI, and the notice, audit, and record-keeping duties fall on the employer.
How: Read plain-language summaries of Local Law 144, the Illinois amendments, and Colorado's AI Act, then ask each vendor what is automated, what the bias audit found, and how candidates are notified.
Testing output for bias
Why: A job description, rubric, or review summary can carry bias in wording you would never notice on one read.
How: Run the same prompt twice with swapped names, genders, or ages and compare. Ask the assistant to flag language that could disadvantage a protected group.
Tools worth knowing
Microsoft Copilot
AI inside Word, Excel, Outlook, and Teams, with your company's data protections.
ChatGPT
The general-purpose AI assistant most of your coworkers already use.
Claude
A careful writing and analysis assistant that shines on long documents.
Gemini
Google's assistant, strongest when your work already lives in Google Workspace.
NotebookLM
A research notebook that only answers from the sources you give it, with citations.
Perplexity
An answer engine that cites its sources, built for research rather than chat.
Cautions for hr professionals
Title VII, the ADA, and the ADEA apply whether a person or a tool made the decision, and disparate impact from an AI screen is the employer's liability even when a vendor built it (the Mobley v. Workday litigation has tested vendor liability since 2023). Keep a human decision-maker in every hire, promotion, and termination, ask vendors for adverse-impact analyses, and use the four-fifths rule from the Uniform Guidelines on Employee Selection Procedures as an early warning.
New York City's Local Law 144 requires an annual independent bias audit of automated employment decision tools, a public summary of results, and notice to NYC candidates at least ten business days before use. Illinois requires consent for AI analysis of video interviews, and its Human Rights Act amendments (effective January 1, 2026) bar AI with a discriminatory effect and require notice when AI informs employment decisions. Colorado's AI Act treats hiring tools as high-risk; its effective date has moved, so check the current one. California's civil-rights regulations now cover automated decision systems, and the EU AI Act classifies hiring AI as high-risk. Confirm each with counsel.
Resumes, personnel files, leave records, and accommodation requests are personal data. California's CPRA covers employee and applicant data, GDPR covers EU staff and candidates, and the ADA requires medical information to sit in separate confidential files. Never paste identifiable employee data into a consumer AI tool unless your organization has approved it and turned off training and retention; anonymize anyway. Prompts and outputs are also records, and may be discoverable in litigation.
Your 30-day plan
- Week 1: Confirm which AI tools are approved and how their data settings work. Inventory where AI already touches employment decisions (ATS ranking, assessments, video interviews) and write your red list.
- Week 1: Build prompt templates for three recurring writing tasks (policy updates, announcements, job descriptions) and use them on real work.
- Week 2: Run an anonymized survey or exit-interview analysis and compare the themes with what you would have said from memory.
- Week 3: Pilot a document-grounded assistant on your handbook with twenty real questions, and send your ATS and assessment vendors the bias, notice, and retention questions.
- Week 4: Draft HR-team guidelines for AI use, brief the managers you support, and total the hours saved.
Frequently asked questions
Will AI replace HR professionals?
Is it legal to use AI to screen job applicants?
Can I paste employee information into ChatGPT?
Which AI tool is best for HR?
Terms used on this page
Related roles
- AI for RecruitersAI takes the typing out of recruiting: job posts, search strings, outreach, and interview guides in minutes. The judgment call is still yours, and the law increasingly says so.
- AI for Operations ManagersOperations managers live in SOPs, incident reports, staffing plans and spreadsheets, which is the material AI handles best. Here is how to use it to get your week back, and where the line is for safety and people decisions.
- AI for Small Business OwnersAI answers the one-star review calmly, drafts the month of posts, turns your voice memo into an SOP, and preps the questions for your CPA. You still make the calls, sign the checks, and own what goes out under your name.
- AI for Executive AssistantsInbox triage, drafts in your executive's voice, meeting briefs, itineraries, and minutes: AI handles the first pass on all of it. The catch is confidentiality, and this page takes it seriously.