A short policy can give teams confidence by explaining where AI can help, who decides and how outputs are reviewed. This template is a starting point, not legal advice or a policy automatically suitable for every company.

A template to adapt

1. Purpose

“We use AI to support writing, organisation and analysis. People remain responsible for decisions and content used.”

Identify covered work and concrete objectives. A useful policy explains permitted activities, not just restrictions.

2. Tools and accounts

List approved tools, plans and account types. Name the person responsible for assessing changes. Review new connectors or features before giving them company data.

3. Information

Define which data is allowed in each tool. Start exercises with fictional or properly anonymised data. Confidential or personal information requires authorisation and an assessment of applicable processing; removing a name may not adequately anonymise a document.

4. Review and decisions

Identify who checks facts, calculations, recipients and external content. Define actions needing approval, such as sending a proposal, changing a record or publishing information.

5. Questions and incidents

Provide an internal contact for questions, unexpected outputs and inappropriate data sharing. Explain what to report without asking people to disclose sensitive information again in an unsuitable channel.

6. Follow-up

Record the version, review date and responsible person. Reassess the policy when tools, data or processes change.

Put it into practice

Use two or three examples from the team’s work. A generic example could be drafting a reply from public information, checking its source and approving the text before sending.

Allow questions and improve the policy with experience. Understandable rules are more useful than a long document people cannot apply.

Data protection, contracts and applicable legal requirements should be assessed with the competent people before processing personal information or integrating systems.