AI policy for your team: what to include (with template)
Clear ground rules for ChatGPT, Copilot and similar tools: the seven building blocks of an internal AI policy for small teams, with a free template.
Marc Herzog Published on 3 min read
In many businesses AI is already in use, often without anyone having decided on it: one colleague drafts emails with ChatGPT, another has meeting notes summarised. That is basically a good thing. Without shared ground rules, however, uncertainty and risks quickly arise, for example when customer data is pasted into a freely available tool.
A short internal AI policy creates clarity. It does not need to be long: for a small team, one or two pages are often enough.
Why it is worth it
- Certainty for the team: everyone knows what is allowed, and nobody experiments in secret.
- Data protection: confidential information stays where it belongs.
- Better results: approved tools and shared rules ensure reliable quality.
- Evidence: a policy is a good building block of the AI literacy measures the EU AI Act expects of businesses.
The seven building blocks
1. Purpose and scope
State in two sentences why the policy exists and who it applies to, for example all employees, trainees and freelancers.
2. Approved tools
List which AI tools may be used, with which access (company account or personal account) and for what. Anything not on the list must be requested from the responsible person before use.
3. Data: what may go in and what may not
This is the most important part. A simple traffic light helps:
| Light | Examples |
|---|---|
| Green: may be entered | publicly available information, general questions, your own texts without personal data or trade secrets |
| Amber: only in approved company tools | internal processes, anonymised examples, drafts without names |
| Red: never enter | personal data of customers and staff, credentials, contracts, financial figures, trade secrets |
Which data you may enter where also depends on your contracts with the tool provider. Agree the traffic light with your data protection officer.
4. Check results
AI can be wrong while sounding very convincing. Specify that a human checks results before use, especially figures, facts, quotations and legal statements. The person who uses a result always remains responsible for it.
5. Labelling
Decide when AI assistance is disclosed, for example for a chatbot on your website, AI-generated images or texts for the public. The AI Act requires this in certain cases.
6. Responsibility and contact person
Name a person who answers questions, reviews new tools and keeps the list up to date. In small businesses this is often the owner.
7. Training and updates
Record how the team is trained and when the policy is reviewed, for example once a year and whenever a new tool is added.
How to introduce the policy
- Take stock: ask openly who uses which AI tools for what. Without blame, otherwise you will not hear the truth.
- Draft: fill in the template. An hour is usually enough for a first draft.
- Discuss: go through the rules with your team and answer questions. Rules people understand are more likely to be followed.
- Publish: store the policy where everyone can find it and ask for confirmation that it has been read.
- Follow up: after a few weeks, check whether the rules work in practice and adjust them.
The template
The free template at the top of this page contains all seven building blocks with sample wording and fields to fill in. It is deliberately short and meant as a starting point. Adapt it to your business.
If you would rather develop the policy together with your team: in my workshop we create a draft of your AI ground rules directly with the people involved.
This article provides general information and is not legal advice. For binding advice on your specific case, please consult a lawyer.