Safety and privacy for AI at work

Lesson 5 of 6

Writing a one-page AI policy

A genuinely doable, one-page policy a small business can write this week and actually use, not a twenty-page document nobody reads.

The moment more than one person in your business uses AI, someone will do something with it you would not have chosen. Not out of carelessness, just because nobody ever told them where the line was. The fix is not a policy the size of a phone book. It is one page, in plain English, that people can actually remember, and here is how to write yours.

Section 1: What AI is fine for

Start with permission, not prohibition, or the whole page reads like a telling-off. List the everyday jobs where using AI is welcome: drafting emails, tidying up your own writing, summarising long documents, brainstorming, getting a rough first version of something you will finish yourself.

Example wording: “You are encouraged to use AI to draft, summarise, tidy and brainstorm. It is a helper, not the final say.”

Section 2: What AI must never be used for

Now the hard lines, kept few and clear. Never send anything straight to a customer or supplier without a person reading it first. Never treat an AI answer as legal, financial or safety advice. Never let it make a final decision about a person, hiring or a complaint, on its own.

Example wording: “AI never gets the final word on money, people, safety, or anything that leaves the building unread.”

Section 3: What must never be pasted in

This is the section that protects you most, so make it concrete rather than vague. People cannot follow “be careful with data”. They can follow a list: no customer personal details, no card or bank information, no passwords, no confidential documents, no staff records. You built exactly this list in an earlier lesson.

Example wording: “If you would not put it on a postcard, do not paste it in.”

Section 4: Who to ask

Most mistakes happen in the grey area, when someone is unsure and does not want to look daft by asking. Fix that by naming one person on the page as the person to check with, and making it plain that asking is never a nuisance.

Example wording: “Not sure if something is OK? Ask [name] first. Asking is never a problem. Guessing wrong is.”

The golden rule goes at the top

Above all four sections, put the line that matters most: AI drafts, a person checks. Nothing reaches a customer, a supplier or the taxman without a human eye on it first. If your team remembers only one sentence from the whole page, make it that one.

Share it properly

A policy sitting unread in a folder does nothing. Walk your team through it in five minutes, pin it up or drop it in the group chat, mention it when someone new joins. Revisit it every few months, because the tools change and your one page should keep up with them.

That is the whole thing: four short sections and a golden rule, no jargon, nothing a lawyer needed to write. It will not cover every situation imaginable, and it is not meant to. It is meant to keep an ordinary team sensible on an ordinary day, which is what actually prevents the real messes.

Next, the final lesson: how to make all of this a habit rather than a one-off training session, and what to do if something does go wrong.

Ready to write yours? Our full template lives at A one-page AI policy for small teams, ready to copy and adapt.

Check your understanding

What is the single most important line to put at the top of a one-page AI policy?

Answer the check above to continue.