A Small-Team AI Policy People Will Actually Follow

Start with allowed tasks

People need to know what is clearly allowed. For many teams, safe starting uses include drafting public blog outlines, rewriting non-sensitive emails, summarizing public information, creating internal checklists, and brainstorming names or ideas. Listing allowed tasks reduces confusion.

Name prohibited data

Be specific about what cannot be pasted into general AI tools: passwords, API keys, customer personal data, financial records, contracts, unreleased strategy, employee records, medical information, legal disputes, and source code unless approved. "Be careful" is not a policy. Clear examples are.

Create an approval path

Some use cases are not obviously safe or unsafe. Give people a simple way to ask. For example: "If the prompt includes client data, confidential business information, regulated data, or a customer-facing promise, ask the team lead before using AI."

Require human review

Every AI-assisted output that leaves the team should be reviewed by a person. The reviewer checks accuracy, tone, missing context, sensitive information, and whether the output makes promises the team cannot keep.

Keep the policy short

A useful small-team policy can fit on one page: allowed uses, prohibited data, approval path, review requirements, tool list, and owner. Review it monthly as tools and business needs change.

How to apply this guide

Use this guide for repeated workplace tasks where clarity, review, and privacy matter more than speed alone. The point is not to automate everything. The point is to let AI handle a draft or structure while a person keeps the business context.

• Write the task in one sentence before opening an AI tool.
• Decide which parts need human review: define allowed use, protect sensitive data, require review.
• Remove private or unnecessary context before prompting.
• Check whether the final output changes a fact, promise, number, date, or decision.

Before using the output, check whether it changes a commitment, exposes private details, adds facts that were not in the source, or removes a nuance that your team actually needs.

A safer prompt to try

Use this starter prompt when you want help with the idea in this guide but still want the model to show its limits.