Press Enter to search  ·  Esc to close

HR Workflow

How to Draft an HR Policy with Claude

You'll finish with a complete, professionally structured HR policy document — even if you've never written one before, even if the company has no existing HR documentation, and even if you're not sure what the legal requirements are. Paste your requirements, run one prompt, refine twice. That's the whole thing.

Free 30 min Freelancers & HR consultants

Have this reviewed before you use it. Claude can draft a clear, well-structured policy — it cannot verify legal compliance. Before adopting any HR policy officially, have it reviewed by a qualified HR consultant or employment lawyer in your jurisdiction. This is especially true for policies covering disciplinary procedures, termination, data privacy, and equal opportunity.

TL;DR

  • You'll have a complete policy draft in under two minutes — from a handful of bullet points.
  • Free plan works fine; one conversation handles a full policy including follow-up edits.
  • Claude writes the structure and language — you supply the rules and sign off on the content.
  • What used to take half a day takes 30 minutes, including the review pass.

Before you start

  • A Claude.ai account — free plan handles a full policy draft without issue.
  • A clear sense of what the policy needs to cover: which employees it applies to, which situations or behaviours it governs, and any specific rules the business owner has already decided on. Even rough bullet points work — don't clean them up first.
  • The country or jurisdiction where the policy will apply. Employment law varies significantly — Claude needs this to frame the language correctly, even if you're handling the legal review separately.
  • Any existing reference material: a previous policy, an industry template, or a contract clause you want reflected. Not required, but it shortens the editing phase significantly.

How to do it — step by step

Define exactly what the policy needs to cover

Start by writing down — before opening Claude — the specific rules you want the policy to enforce. Not "something about remote work" but: who can work remotely, how many days per week, whether manager approval is required, how equipment is handled, and what happens when someone doesn't follow the rules. Spend an extra minute listing the exact rules you want the policy to cover, because vague input produces vague output and you'll spend three rounds of edits fixing something that five minutes of thinking would have prevented. If you're working with a client who hasn't thought this through yet, use the HR Prompt Builder — it has a scoping prompt designed to pull these decisions out of a business owner in a structured conversation before you touch the policy itself. Once you have the rules written down, note the policy type (Remote Work, Disciplinary, Leave, Data Privacy, etc.) and which employees it applies to — all staff, full-time only, managers. That's your input.

Collect any reference material you have

At this point, pull together anything that already exists: a policy from a previous client in the same industry, a government template for your jurisdiction, a clause from an employment contract, or a competitor's publicly available handbook. You're not plagiarising — you're giving Claude anchoring material so the output matches the register and format your client expects. Don't worry if the reference is outdated or from a different company size; Claude will adapt it. If the company has no existing HR documentation at all, skip this step and move straight to the prompt — Claude can work from your scoped requirements alone.

Run the prompt and get a first draft

Now open Claude and paste the prompt below. Fill in the green placeholders — policy type, jurisdiction, who it applies to, your list of rules — then paste any reference material at the bottom. Send it. Claude returns a structured draft with standard sections (Purpose, Scope, Policy Statement, Procedures, Responsibilities, Review Date) in plain, professional language, usually in under 90 seconds. Don't try to perfect the prompt before sending. A rough first pass gives you something concrete to react to, which is always faster than trying to get the prompt exactly right before you've seen any output. The goal at this stage is a working skeleton you can read through and react to — not a document you'd hand to a client today.

Refine for the company's actual context

This is where most of the real work happens. Read the draft once and flag anything that doesn't fit — language that's too formal for a 10-person startup, a procedure that assumes an HR department that doesn't exist, a responsibility assigned to a "line manager" when this company doesn't use that title. Then go back into Claude and fix them one section at a time. "Rewrite the Disciplinary Procedure section to reflect a three-step verbal warning, written warning, final written warning process" is a better follow-up request than "make it more practical." The AI for HR Teams hub has specific examples of how to phrase these refinement requests for different policy types. Expect two to three rounds of editing. That's normal for a policy of any complexity — Claude gets you to a strong draft quickly, but the fit to a specific company's culture and language takes iteration.

Add metadata and flag for legal review

Before the draft leaves your desk, fill in the policy metadata: version number, effective date, review date, owner (the role responsible for maintaining the policy, not the person). Then add the legal review disclaimer — see the note at the top of this page. Send it to the relevant HR or legal reviewer with a clear note about which sections you're most uncertain about. Done.

The prompt

You are helping me draft a professional HR policy document for a company.

Context:
- Policy type: [e.g. Remote Work Policy / Disciplinary Policy / Annual Leave Policy / Data Privacy Policy]
- Company size and type: [e.g. 25-person marketing agency / 8-person e-commerce startup]
- Jurisdiction: [e.g. United Kingdom / California, USA / Australia]
- Who this policy applies to: [e.g. all permanent employees / full-time and part-time staff / managers only]
- Tone: [e.g. professional and formal / plain language and approachable / legally precise]

Rules and requirements this policy must include:
[List every rule, limit, or procedure the policy needs to cover — bullet points are fine. Be specific: "Employees may work remotely up to 3 days per week with manager approval" rather than "something about remote work limits."]

Reference material (optional):
[Paste any existing policy, template, or relevant contract clause here. Leave blank if you don't have one.]

Using the rules and requirements above, draft a complete HR policy document with the following structure:
1. Purpose
2. Scope
3. Policy Statement
4. Procedures (step-by-step where relevant)
5. Roles and Responsibilities
6. Non-compliance
7. Review and Updates

Write in clear, professional language appropriate for the tone I specified. Do not add rules or requirements that are not in my input above. If a standard section has no relevant content based on my requirements, say so briefly rather than padding it with generic text.

Format it as a complete policy document I can refine and send for legal review.

What the output looks like

Example output — Remote Work Policy

1. Purpose

This policy sets out the conditions under which employees of Alderton Digital Ltd may work remotely, and the responsibilities of both the employee and their line manager in supporting effective remote working arrangements.

2. Scope

This policy applies to all permanent full-time and part-time employees of Alderton Digital Ltd based in the United Kingdom. It does not apply to contractors, freelancers, or employees on fixed-term contracts of less than three months.

3. Policy Statement

Alderton Digital supports flexible working and recognises that remote work can benefit both employees and the business. Employees may work remotely for up to three days per week, subject to manager approval and the operational requirements of their role. Remote working is a privilege, not an entitlement, and may be modified or withdrawn if performance or collaboration is adversely affected.

4. Procedures

Employees wishing to establish a regular remote working arrangement must submit a written request to their line manager at least five working days in advance. The manager will confirm approval or request a discussion within three working days. Ad hoc remote working requests may be made informally on the day, subject to manager discretion.

Claude vs Gemini for this workflow

Dimension Claude Gemini
Following a precise list of rules Better Good
Professional, legally appropriate tone Better Good
Avoiding invented content Better Good
Integration with Google Docs Good Better
Speed on free plan Good Better
Overall recommendation Recommended Solid alternative

How to customise this workflow

For small business owners writing their first policy

Add one line to the prompt: "Assume this company has never had a formal HR policy before — write in plain language that any employee can understand without HR training." Then, before filling in the rules, ask Claude in a separate message: "What are the minimum things a Remote Work Policy for a 10-person UK company should cover?" It'll give you a checklist in 20 seconds that you can use to fill in the requirements field. The AI for HR Teams hub has more guides written specifically for non-specialists doing HR work without a dedicated team.

For HR consultants managing multiple clients

Don't rewrite the prompt for every engagement — save a base version per policy type in your notes tool, then swap in the client's company name, size, jurisdiction, and specific rules before pasting. That drops the setup from ten minutes to two. The HR Prompt Builder is built exactly for this: configure the prompt once per policy type, store it, and reuse it without starting from scratch each time.

For policies that need to reference existing documentation

Skip this if you're starting from scratch. But if the client already has an employment contract or employee handbook, paste the relevant sections into the reference material field and add this instruction: "Ensure this policy is consistent with the contract clauses I've included — flag any conflicts rather than silently overriding them." Claude will surface inconsistencies as notes in the draft, which saves you from discovering them during the legal review.

For sensitive policy types (disciplinary, termination, harassment)

Before moving on to the prompt, add one line to the context section: "This policy may be used in formal disciplinary proceedings — ensure the language is precise, unambiguous, and procedurally sound." Claude tends to soften language in sensitive areas by default; this instruction tells it to prioritise clarity over diplomacy. Then flag the draft explicitly for legal review before it goes anywhere near an employee — these are the policy types where jurisdiction-specific employment law matters most.

Frequently asked questions

Can Claude draft a legally compliant HR policy?
Claude can draft a structurally sound, professionally written policy — but it cannot verify legal compliance. Employment law varies by jurisdiction, changes regularly, and often depends on company-specific factors that aren't in your prompt. Use Claude to produce the draft, then have a qualified HR consultant or employment lawyer review it before it becomes official. That review is faster and cheaper when you're handing them a clean draft rather than a blank page.
What if I don't know what the policy should say?
Ask Claude before running the main prompt. In a separate message, type: "What does a [policy type] policy for a [company size] company in [jurisdiction] typically need to cover?" Claude will give you a starting checklist you can use to fill in the requirements field. Don't worry if you're not sure what the legal requirements are — that's what the legal review step is for. Your job at this stage is to capture the business rules the company has already decided on; Claude handles the structure and language.
Is it safe to include company-specific details in my prompt?
Anthropic doesn't use your conversations to train Claude by default. That said, check your client's data handling policy before pasting any identifiable company information into an external tool — some organisations have blanket rules about this. A safe workaround: use a generic company name (e.g. "Company A") in the prompt and replace it with the real name when you paste the final draft into your document. The output quality is identical.
How do I handle policies that need to apply across multiple countries?
Draft one version per jurisdiction rather than trying to write a single cross-border policy — the differences in employment law are usually too significant to paper over with one document. Run the prompt once per country, changing the jurisdiction field each time. Then, in your follow-up request, ask: "Flag any sections that are likely to differ significantly between [Country A] and [Country B] — I'll need to check those with local counsel." The HR Workflows hub has a multi-jurisdiction checklist that helps you track which versions are at which stage of review.