The One-Page AI Policy Your Office Team Actually Needs (With a Template You Can Copy)

If your company runs on Google Workspace or Microsoft 365, your office team can already use Gemini or Copilot, and some of them probably do. A one-page AI policy tells them what is safe to paste into it, who checks the result, and who to ask when they are unsure. That is all most small manufacturers and distributors need.

I hear the same thing in most offices I work with. The owner says, “I think a few people use it, but I couldn’t tell you how.” The office manager says, “I’m not sure what we’re allowed to put in there.” Nobody is doing anything reckless. Nobody has been told where the lines are.

This guide gives you the one page, a copyable template, and a simple way to roll it out in a week.

Why a policy comes before the training

When people have a powerful tool and no rules, they guess. One person pastes a customer’s full email thread with account details. Another pastes only the question. A third decides AI is off limits and quietly keeps typing everything by hand.

A short policy fixes three things at once:

• Safety. People know what never goes into a prompt.
• Accuracy. Someone is named as responsible for checking what goes out the door.
• Confidence. Hesitant staff get permission to try the tool on real work.

That last one gets overlooked. A clear policy doesn’t only restrict people. It tells the careful ones that using AI on a quote follow-up is allowed.

What goes on the one page

Keep it to six short sections. Each one answers a question an employee would otherwise ask you.

1. Approved tools

Name them. “Gemini in our Google Workspace account” or “Copilot in our Microsoft 365 account.” Say that work tasks go through these and not through personal accounts of free tools. On most current business plans, Google and Microsoft state that your data stays inside your organization and isn’t used to train their models. Check the terms for your own plan before you write that sentence.

2. What never goes into a prompt

This is the section people remember. Keep it as a short list:

• Payment card or bank details
• Passwords, logins and access codes
• Employee personal data such as Social Security numbers, medical details or pay
• Another customer’s private information, when you are writing to a different customer
• Anything covered by a contract or NDA that limits sharing

3. What is fine to use it for

Be just as specific here. Quote follow-ups, order-status replies, summarizing a long email thread, cleaning up a job post, drafting onboarding instructions. Naming real tasks tells people the tool is meant to be used.

4. Who checks the output

Every AI draft that reaches a customer, vendor or candidate gets read by a person first. The person who sends it owns it. This one line prevents most of the embarrassing mistakes.

5. Customer-facing rules

Decide whether you want people to tell customers when AI helped write something. Most offices say that routine emails need no label, but anything that sounds like a commitment on price, lead time or quality gets a human check against the real numbers.

6. Who to ask

One name, one email address. Add a line that says questions are welcome and nobody gets in trouble for asking.

A template you can copy

Paste this into a document, change the bracketed parts and put your name at the top.

[Company] AI Use Policy
Owner: [Name, role, email]    Last reviewed: [Date]

1. Approved tools. We use [Gemini in Google Workspace / Copilot in
   Microsoft 365] with our company account. Do not use personal AI
   accounts for company work.

2. Never put these in a prompt: payment or bank details, passwords,
   employee personal data, another customer's private information,
   or anything an NDA or contract says we can't share.

3. Good uses: drafting emails and replies, summarizing threads,
   cleaning up job posts and SOPs, preparing quote follow-ups,
   turning notes into checklists.

4. Review before sending. A person reads every AI draft before it
   goes to a customer, vendor or candidate. The sender is responsible
   for what it says.

5. Check the numbers. Prices, lead times, quantities and dates come
   from our systems, never from the AI's memory.

6. Questions. Ask [Name] at [email]. Asking is always okay.

That is the whole policy. It is under 150 words, and it covers the situations that come up in a real office week.

Roll it out in one week

Writing the page is the easy half. Making it stick takes a few small steps.

Monday and Tuesday: draft and review

Fill in the template and have the owner or operations lead read it. Remove anything that doesn’t match how your office works. If a rule would be ignored, cut it.

Wednesday: walk the team through it

Fifteen minutes, in person or on a call. Don’t just read it out. Show one real example: take a customer email, paste only the question and the facts, and let Gemini or Copilot draft the reply. Then show what you would not paste. If you want a worked example, the guides on quote follow-up emails with Gemini and answering “where’s my order?” emails use exactly this pattern.

Thursday: collect acknowledgements

Ask each person to reply “read and understood” by email. It takes a minute and gives you a record.

Following week: check in

Ask the office manager three questions. What did people use it for? What was unclear? Did anything come up the policy didn’t cover? Adjust the page and re-send it.

Mistakes I see offices make

Making it too long. If it doesn’t fit on one page, the details won’t be remembered when someone is mid-email at 4:45 on a Friday.

Making it too vague. “Use AI responsibly” tells nobody anything. “Never paste payment details or passwords” does.

Having no owner. A policy with no name on it is a policy nobody updates. Put one person’s name at the top and a review date six months out.

A fourth mistake is banning the tool outright. People still use it, on personal accounts, where you have no say at all. An approved option with clear rules is safer than a ban that gets ignored.

What a policy doesn’t do

A policy sets boundaries. It doesn’t teach anyone to write a good prompt, and it doesn’t tell you which task to start with. That part is training, and it works best when it is tied to one real task your office does every week.

I include a one-page policy in the free AI Quick Win Kit for exactly that reason. It sits next to a custom agent for your team’s most annoying task, a prompt pack for your office, and a check-in after a week where you measure the before and after yourself. If you’d like that for your team, the box below has the details.

Frequently asked questions