Your team already uses AI. Do you have anything in writing?

How we use AI at [COMPANY NAME]

Owner: [RESPONSIBLE PERSON] · Date: [DATE] · Next review: [REVIEW DATE]

This is how we use AI tools (ChatGPT, Copilot, Claude, Gemini and the like) to save time without putting the company, our customers or ourselves at risk. It takes two minutes to read.

1. What we use AI for, and what we don't

Yes: drafting and rewriting text, summarising, organising ideas, translating, preparing first drafts and explaining things.

Always with a check: pulling data from a document into a table, comparing versions, technical drafts.

No: making decisions about people (hiring, firing, appraisals), trusting a figure, a rule or a quote without checking it, or sending anything to a customer before a person has read it.

2. Approved tools and accounts

We only use [APPROVED TOOLS], always through the company account. No personal or free accounts for work, and no new tool gets installed without asking [RESPONSIBLE PERSON] first. Need another one? Ask and we will look at it together.

3. Data that never goes in

Never paste or upload any of the following into an AI tool that is not the approved one:

• personal data about customers, suppliers or colleagues (names with ID numbers, phone numbers, emails, payslips, health information);
• contracts, quotes and prices that are not public;
• passwords, access keys or codes;
• internal reports, customer lists or financial information.

When in doubt, the answer is no, and you ask. If you only need a summary, strip out the names and details and ask anyway: the AI works just as well.

4. Check before you use it

Sounding right is not the same as being right. Before using what an AI tells you, run through these:

1. Source: where does it come from, and does it exist?
2. Figures: have I worked them out again myself?
3. Dates: are they current?
4. Names: do these people and companies actually exist?
5. Rules and regulations: have I checked the official source?
6. Sense: does it fit with what I know?

Asking the same AI whether it got it right is not a check. If it matters, a second person reviews it.

5. Accountability

The AI suggests; the person who uses it answers for it. Whatever we send, sign or publish is the responsibility of whoever sends, signs or publishes it, even if a tool wrote the first draft.

6. When something goes wrong

If you have uploaded something you should not have, spotted a wrong answer that has already been used, or seen someone use a tool that is not approved, tell [RESPONSIBLE PERSON] the same day. Speaking up early is not a telling-off: it stops a slip becoming a real problem.

7. Training

Everyone takes the free AI literacy training (a short video and a quiz) at aiprocessia.com/en/ai-literacy-certificate/. It earns a certificate of completion, not an official qualification. We review this policy at least once a year, and whenever our tools change.

Read and agreed

I have read this policy and I will follow it.

Name: ______________________ Signature: ______________________ Date: ____________

---

A starting template from Softnet Sistemas and AIPROCESSIA. Adapt it to your business. It is not legal advice and does not, on its own, guarantee compliance with any regulation.