HR, people and recruiting
Write job descriptions and structured interview kits, screen with bias guardrails inside today's legal limits, build onboarding plans, answer policy questions from your handbook, write reviews with care and protect employee data.
For HR business partners, recruiters and talent acquisition teams, people operations and HR leaders
Hire and support people faster, with fairness and privacy built into every step
- A job description and a structured interview kit with scorecards in under 30 minutes
- Every candidate for a role assessed by a person on the same job-related criteria
- Policy questions answered from your handbook with the section quoted, or escalated
- A 30-60-90 onboarding plan ready before each new hire's first day
- Employee data kept out of every tool that lacks approved data terms
On this page
- Where AI helps in this job
- Start in 15 minutes
- The tools people in this role use
- Members only: Workflow: job description to interview kit
- Members only: Workflow: screening with bias guardrails
- Members only: Workflow: onboarding plans that start before day one
- Members only: Workflow: a grounded policy Q and A assistant
- Members only: Performance reviews, written with care
- Members only: People analytics without exposing individuals
- Members only: Employee-data privacy
- Members only: A skill for HR
- Members only: Guardrails and hooks
- Members only: Routines that run without you
- Members only: Prompt ladder for HR
- Members only: Measure it
- Members only: Risks and what not to automate
HR work runs on cycles that repeat: requisitions, interviews, onboarding, policy questions, review season and the monthly people report. AI drafts all of that quickly. It is unreliable at judging people, and it can repeat bias from its training data or from your past hiring decisions. Lawmakers treat hiring as a high-risk use for exactly that reason. The rule on this page is short: AI prepares, a person decides, and employee data goes only into tools your company approved.
Where AI helps in this job#
| Task | What AI does well | What you still own |
|---|---|---|
| Job descriptions | Drafts from intake notes, removes jargon and gendered wording, questions requirements nobody needs | Level, pay range and the final requirements |
| Structured interview kits | Writes competency questions, probes and 1-3-5 scoring anchors tied to the job | Which competencies matter, interviewer calibration |
| Screening support | Lists evidence for each job-related criterion, with quotes from the application | Every decision about a candidate |
| Candidate communications | Drafts outreach, scheduling and rejection emails in your tone | Sending them, and anything sensitive |
| Onboarding plans | Builds 30-60-90 plans, first-week calendars and checklists | Introductions, equipment and access |
| Policy questions | Answers from your handbook with the section quoted | Exceptions and anything about a named person |
| Performance reviews | Turns a manager's notes into specific, behaviour-based wording | The rating, the judgment and the conversation |
| People analytics | Computes attrition, headcount and time to hire from exports with code | What to do about the numbers |
| Policy drafting | First drafts and FAQs from the requirements you supply | Legal review and consultation with employees |
AI does not know your local employment law or your policies unless you paste them in, and it will answer confidently either way. It cannot verify a candidate's experience or identity. It reads a CV the way its training taught it to, which can reward the wrong signals: school names, writing style, employment gaps. Treat its output on people as a draft for a person to check, never as an assessment.
Start in 15 minutes#
These work in Claude, ChatGPT, Gemini or Microsoft Copilot (formerly Microsoft 365 Copilot). Use your company account.
Prompt
Make a job description clearer and more inclusive
Rewrite this job description so someone outside our company understands it and every qualified person feels able to apply. Remove jargon, gendered wording and requirements that are not needed to do the job. List every change you made with the reason. Do not invent pay, benefits or duties. [PASTE JOB DESCRIPTION]
Good output gives you a list of changes you can accept or reject one by one. Check that every requirement left in the description maps to a real task in the role.
Prompt
Structured interview questions for one competency
Write 3 behavioural interview questions that test [COMPETENCY, FOR EXAMPLE STAKEHOLDER MANAGEMENT] for a [ROLE] at [LEVEL]. For each question, add two follow-up probes and a scoring guide that says what a 1, a 3 and a 5 answer sound like. Use this format: Question: Tell me about a time you had to change a plan after it was agreed. Probes: What made you change it? Who did you tell first? 1: Describes the change only, with no reasoning or communication. 3: Explains the reason and tells the people affected. 5: Explains the trade-off, tells people early and checks the result afterwards. Ask only about past behaviour at work. Do not ask about age, family, health, religion, nationality or any other personal characteristic.
Good questions are ones any qualified candidate could answer from their own work. Ask every candidate for the role the same questions, in the same order.
Prompt
Answer a policy question from the handbook only
Answer the employee question below using only the handbook text I paste. Quote the section title and the exact sentence you rely on. If the handbook does not answer the question, reply "The handbook does not cover this" and suggest who in HR to ask. Do not guess and do not use general knowledge. Question: [QUESTION] Handbook text: [PASTE THE RELEVANT POLICY SECTIONS]
Check that the quoted sentence really appears in your handbook. A good answer is short, quotes the source and admits when the handbook is silent.
Warning
Candidate and employee data needs approved tools. Do not paste CVs, pay, health information, grievances, investigation notes or ID numbers into a personal chat account. Use your company plan and check its data terms. Anthropic, for example, does not train on inputs or outputs from its commercial products (Claude Team, Enterprise and the API) by default. Remove identifying details whenever the task does not need them.
The tools people in this role use#
- A general assistant on your company plan. Claude, ChatGPT, Gemini or Microsoft Copilot for drafting, analysis and handbook questions.
- Anthropic's human-resources plugin. Free and open source, for Claude Cowork and Claude Code. Its skills include draft-offer, interview-prep, onboarding, performance-review, policy-lookup, comp-analysis, people-report, org-planning and recruiting-pipeline. It works with details you paste in, or with HRIS and ATS connectors you add; its README names Workday, BambooHR, Rippling, Greenhouse, Lever and Ashby as examples. The plugins guide shows how to install and adapt one.
- Policy assistants. Claude Projects hold handbook files and instructions on every plan, and Team and Enterprise plans can share a project with colleagues. Microsoft Copilot's Agent Builder can use SharePoint content as knowledge. Gemini Gems take uploaded files as knowledge.
- AI features in your ATS or HRIS. Ask the vendor for its bias audit results, model documentation and data retention terms before you switch anything on. In New York City, a bias audit is a legal requirement before you use an automated employment decision tool.
Workflow: job description to interview kit
Members only below this line
Get the full HR, people and recruiting playbook free
What you get on this page:
- 4 step-by-step workflows
- 13 prompts
- 1 skill
- 1 hook
- 2 routines
Also in the members part
- Workflow: screening with bias guardrails
- Workflow: onboarding plans that start before day one
- Workflow: a grounded policy Q and A assistant
- Performance reviews, written with care
- People analytics without exposing individuals
- Employee-data privacy
- A skill for HR
- Guardrails and hooks
- Routines that run without you
- Prompt ladder for HR
- Measure it
- Risks and what not to automate
Sources
Checked against these on 13 Oct 2026.
- European Commission: AI Act application timeline (opens in a new tab)digital-strategy.ec.europa.eu
- EUR-Lex: Regulation (EU) 2024/1689, the AI Act (Annex III, Articles 5 and 26) (opens in a new tab)eur-lex.europa.eu
- EUR-Lex: Regulation (EU) 2026/1744, the Digital Omnibus on AI (opens in a new tab)eur-lex.europa.eu
- NYC Department of Consumer and Worker Protection: Automated employment decision tools (opens in a new tab)nyc.gov
- California Civil Rights Council: regulations on automated-decision systems (June 2025) (opens in a new tab)calcivilrights.ca.gov
- Colorado General Assembly: SB26-189, Automated Decision-Making Technology (opens in a new tab)leg.colorado.gov
- ICO: Thinking of using AI to assist recruitment? Key data protection considerations (November 2024) (opens in a new tab)ico.org.uk
- California Attorney General: CCPA (employee data exemption expired 31 December 2022) (opens in a new tab)oag.ca.gov
- 29 CFR 1607.4(D): the four-fifths rule in the Uniform Guidelines (opens in a new tab)law.cornell.edu
- Anthropic knowledge-work-plugins: human-resources plugin (opens in a new tab)github.com
- Claude Help Center: What are projects? (opens in a new tab)support.claude.com
- Claude Help Center: Use skills in Claude (opens in a new tab)support.claude.com
- Claude Help Center: Use connectors (tool permissions) (opens in a new tab)support.claude.com
- Claude Help Center: Schedule recurring tasks in Claude Cowork (opens in a new tab)support.claude.com
- Microsoft Learn: Agent Builder overview (opens in a new tab)learn.microsoft.com
- Gemini Apps Help: Use Gems (opens in a new tab)support.google.com
- Claude Code docs: Hooks reference (opens in a new tab)code.claude.com
- Claude Code docs: Configure permissions (opens in a new tab)code.claude.com
- Anthropic Privacy Center: commercial products are not used for training by default (opens in a new tab)privacy.claude.com