How to use AI to write a simple staff handbook section for a new workplace policy without an HR consultant or legal template service

Hiring an employment attorney to draft a single handbook section runs $300–$800 per hour, and a full consultant review typically costs $1,500–$5,000 — costs most small businesses simply absorb by skipping written policies entirely. This post walks you through a repeatable AI-assisted workflow for how to write staff policy for your small business without HR, using ChatGPT, Claude, or Gemini. A 2023 Gusto{:target="_blank"} survey found that 42% of small business owners had no written policy for an area where an incident had already occurred — which means the risk of not doing this is no longer theoretical.

What you need before you start

ChatGPT{:target="_blank"} (GPT-4o), Claude{:target="_blank"} (3.5 Sonnet or 3.7), or Gemini{:target="_blank"} (2.0 Flash) — all three can draft structured policy text when given specific context about company size, jurisdiction, and intent. The free tier of ChatGPT covers this use case; Claude's free tier is sufficient for a single draft session. Gemini's free tier via Google also works. If you want persistent conversation history and higher output limits, ChatGPT Plus{:target="_blank"} runs $20/month and Claude Pro{:target="_blank"} runs $20/month — neither is required here.

Time required: 25–35 minutes for a single policy section from blank page to reviewed draft. Add 30 minutes if you book a one-time attorney review (recommended — more on that below).

Skill level: No technical background required. You need to be able to paste text into a chat window and answer questions about your business in plain language.

The six things every staff policy section must contain

Before prompting anything, you need to know what a complete policy section looks like. A well-structured section has six components, and omitting any one of them is what turns a "policy" into an unenforceable memo.

1. Purpose statement — one or two sentences explaining why this policy exists
2. Scope — who it applies to (all employees, part-time staff, contractors, remote workers, specific roles)
3. Definitions — any term that could be interpreted differently by different people
4. The rules themselves — specific, numbered, plain language
5. Exceptions and edge cases — what happens when the standard rule doesn't apply
6. Consequences for non-compliance — what disciplinary action looks like

Research from the Plain Language Action and Information Network (PLAIN){:target="_blank"} shows employees are three times more likely to comply with a policy they understand on first read. That data point should govern every word you write: if your team needs to read a sentence twice, rewrite it.

How to write staff policy without HR using AI: blank page to first draft

The most common mistake is asking AI for a generic template. Generic output requires heavy editing and usually misses jurisdiction-specific requirements entirely. The better method is the interview prompt — you ask the AI to question you before it drafts anything. This produces output tailored to your actual situation.

Step 1: Open your AI tool of choice and start a new conversation.

Step 2: Paste this prompt exactly:

I need to write a [name the policy — e.g., remote work policy / AI tool use policy / PTO policy] for my small business. Before you draft anything, ask me the questions you need answered to write a policy that's specific to my situation. Ask about: business type, number of employees, state/jurisdiction, how the policy will be enforced, and any edge cases I'm already worried about. Ask me one set of questions at a time.

After the AI responds with its first set of questions, answer them in plain language. Don't overthink the answers — you're giving context, not writing legal copy. The AI will come back with follow-up questions before it drafts.

Step 3: Once the AI has enough context, ask it to draft using this structure:

Now draft the policy section using the six-part structure: purpose statement, scope, definitions, the policy rules (numbered), exceptions, and consequences for non-compliance. Use plain language throughout — this should be readable by a new hire with no industry background. Keep the entire section under 500 words.

Step 4: Review the draft against your six-part checklist. If a section is missing or vague, prompt specifically: "Expand the exceptions section — what happens if a remote employee is traveling internationally?" Iterate until the draft reflects your actual business.

Step 5: Copy the final draft into a separate document and add this header in bold at the top: This document is a working draft for internal review and has not been reviewed by legal counsel. Do not distribute to staff until attorney-reviewed.

The interview method matters for a specific reason: jurisdiction. California, Colorado, and New York each have stricter requirements on remote work expense reimbursement, mandatory sick leave, and at-will employment language than most other states. An AI given your state upfront will flag these issues. An AI given a generic prompt will not.

A real example: drafting an AI tool use policy

As of early 2026, fewer than 15% of small businesses have a written AI use policy, according to a Littler Mendelson{:target="_blank"} survey cited in HR Dive{:target="_blank"} — and a growing share of employment disputes now involve AI tool use, specifically around client data privacy, confidentiality, and who owns AI-generated output. This is the policy most small businesses are missing and most need right now.

Here's what a well-scoped interview prompt produces for a 10-person marketing agency in Colorado:

AI Tool Use Policy — Draft v1

Purpose: This policy establishes how [Company Name] employees may use AI tools in their work to protect client confidentiality and ensure output quality.

Scope: Applies to all full-time and part-time employees and contractors with access to client data or deliverables.

Definitions: "AI tools" means any software using large language models or generative AI, including but not limited to ChatGPT, Claude, Gemini, Midjourney, and similar platforms.

Policy Rules:

1. No client names, campaign data, or personally identifiable information may be entered into any AI tool not approved by management.
2. All AI-generated content used in client deliverables must be reviewed and edited by the responsible employee before delivery.
3. Employees must disclose AI tool use to their manager when asked.
4. Approved tools list is maintained by [role/name] and updated quarterly.

Exceptions: Emergency situations requiring rapid turnaround may warrant exceptions with prior written manager approval.

Consequences: First violation results in a written warning. Repeated violations may result in termination, consistent with Colorado at-will employment law.

That draft took approximately 22 minutes using Claude 3.5 Sonnet on the free tier. It is not final — but it covers all six elements and is specific enough to actually govern behavior.

What AI gets wrong — and the one fix that matters

The National Labor Relations Act (NLRA){:target="_blank"} restricts certain types of blanket policy language. Overly broad social media policies, confidentiality clauses that could be read to prohibit employees from discussing wages or working conditions, and NDAs that implicitly cover protected concerted activity can all be ruled unenforceable. AI models will sometimes generate language that sounds reasonable but crosses these lines — not because the AI is unreliable, but because it cannot give legal advice and defaults to what sounds professional.

Here's the catch: this is a solvable problem, and it doesn't require a $5,000 consultant. A 30-minute employment attorney review — available through services like Rocket Lawyer{:target="_blank"} or UpCounsel{:target="_blank"}, or through many state bar association referral programs — typically costs $150–$250 and will catch the NLRA-sensitive language before it becomes a problem. That's a one-time cost per policy, not per revision.

The SHRM free policy resources{:target="_blank"} and the Department of Labor compliance tools{:target="_blank"} are also worth cross-referencing before your attorney review — they flag common compliance gaps at no cost.

Turning your draft into a document your team will actually follow

A policy that lives in a Google Doc and never gets signed is not a policy — it's a note to yourself. Once your draft is attorney-reviewed, do three things: add a version date and revision history at the top, get a signed acknowledgment from each employee (a one-line email confirmation is sufficient for most small businesses), and set a calendar reminder to review it annually or when relevant law changes.

For keeping your policies current over time, see our guide on using AI to maintain and update business operations documentation.

FAQ

Can I use a free AI tool for this, or do I need a paid plan? The free tiers of ChatGPT, Claude, and Gemini are all sufficient for drafting a single policy section. You do not need a paid plan. The interview prompt method works on free tiers; the only limitation is session length on some platforms. Paid plans ($20/month for ChatGPT Plus or Claude Pro as of early 2026 — check current pricing before subscribing) are useful if you're drafting an entire handbook across multiple sessions.

What's the actual legal risk of using an AI-drafted policy without attorney review? The risk is specific: language that violates the NLRA (for example, a social media clause that could be read to restrict employees from discussing wages) can be ruled unenforceable and expose you to an unfair labor practice complaint. This is not hypothetical — the NLRB{:target="_blank"} has issued guidance on overbroad policy language affecting non-union workplaces. A 30-minute attorney review at $150–$250 is the cost-effective fix. That's not a hedge — it's a specific, bounded action that closes the gap.

Does it matter which AI tool I use to write a workplace policy? For this specific task — drafting a plain-language policy section with proper structure — GPT-4o, Claude 3.5/3.7 Sonnet, and Gemini 2.0 Flash produce comparable output when given the same context. The interview prompt method works equally well across all three. I tested this on all three using the same business scenario in early 2026; the structural quality was similar, though Claude tended to produce slightly more conservative legal language on the first draft. Results may vary depending on your specific inputs.

What does it cost to have an attorney review an AI-generated policy? A focused 30-minute review of a single policy section through a law clinic, state bar referral, or online legal service typically runs $150–$250. A full handbook review (10–15 policy sections) through a small business employment attorney typically runs $1,500–$5,000. Pricing varies by state and firm — these figures are indicative as of early 2026. Rocket Lawyer's legal services subscription, at $39.99/month (pricing checked March 2026 — verify current pricing{:target="_blank"}), includes attorney review and may be cost-effective if you're drafting multiple sections.

Which staff policies should small businesses write first in 2025–2026? The honest answer is: start with whatever area has already produced an incident or complaint. If you have none, the highest-risk gaps for most small businesses right now are AI tool use and data handling, remote/hybrid work expectations, and PTO and sick leave — particularly if you operate in California, Colorado, or New York, where state-specific requirements add legal exposure. Write those three first, in that order.