Writing an AI policy
A two-page AI policy that employees actually read.
Download
Version 1.0 · updated 16 November 2026
What you put in
Your organisation, the AI tools in use now, what data employees work with and who will be the point of contact.
What you get back
A first version of an AI policy of no more than two pages, with an example for every rule. Not legal advice: have it reviewed.
Installing
How to add the skill to your AI tool. The download you need is listed next to it.
| Claude | Upload the ZIP via Customize › Skills. Code execution must be switched on. For Claude (.zip) |
|---|---|
| ChatGPT | Import SKILL.md under Skills. In a workspace an admin must switch skills on. SKILL.md |
| Microsoft 365 Copilot (Cowork) | Choose Upload skill and pick the ZIP for Copilot or the .md file. It is stored in your OneDrive. For Copilot (.zip) |
| Gemini | Upload SKILL.md in the web app or the Mac app. SKILL.md |
Checked on 27 October 2026. More explanation and sources in the article on skills. →
The skill
---
name: ai-policy
description: Writes a short, readable AI policy for a team or organisation, covering allowed tools, rules for data, checking AI output and a point of contact. Use this skill when someone asks for an AI policy, AI guidelines, agreements on AI use or AI literacy.
metadata:
owner: "Hello Growth"
version: "1.0"
updated: "2026-11-16"
license: "CC BY 4.0, https://hellogrowth.ai/en/skills"
---
# Writing an AI policy
## Goal
An AI policy of no more than two pages that employees actually read and can follow.
## Input
Ask for: the organisation and number of employees, which AI tools are used now and on which plan,
what kinds of data employees work with (customer data, personal data, financial data), and who will
be the point of contact. Fill in nothing that was not given; write "to be decided".
## Steps
1. Purpose of the policy in two sentences: why, and for whom.
2. Allowed tools: which ones, on which plan, and what is not allowed (for example private accounts for work).
3. Data: what may and may not go into an AI tool, with an example for each kind of data.
4. Checking: AI output is a draft; who checks what before it goes out.
5. AI literacy: how employees learn to work with AI, in a way that suits their role.
6. Point of contact and how often the policy is reviewed.
## Check
- Can the policy be read in ten minutes?
- Does every rule have an example?
- Does it state that this is not legal advice and that the organisation has it reviewed itself?
## Note
The EU AI Act requires organisations that use AI to, among other things, ensure their employees are
sufficiently AI literate. This skill helps write a first version; have the policy reviewed by
whoever is responsible for privacy and compliance in your organisation.
Licence CC BY 4.0: you may use, adapt and share it, crediting Hello Growth.
Adapting it to your team
Put your existing privacy policy or IT rules next to the skill as a file, so the AI policy lines up with them.
Read more
7 October 20266 min readUsing AI safely: how to write an AI policy for your team
An AI policy in five agreements on one page: which tools, which data, who checks and who reports. With what the GDPR asks and a free Word template.
23 October 20267 min readEU AI Act for SMEs: what you must and must not do
The EU AI Act for small and mid-sized businesses in plain language: your role, what to do now, what to skip and the deadlines, with official sources.
5 October 20266 min readAI literacy in your organisation: what the AI Act asks and how to show it
AI literacy is mandatory, even if your team only uses ChatGPT or Copilot. What Article 4 asks after the Omnibus, a four-step approach and what to record.