AI usage policy template
Last updated:
A one-page, editable AI usage policy your organization can copy and adapt: what agents may do, approvals, data to protect, owner and review date.
A one-page policy is easier to follow than a long one. Copy the text below, replace everything in [brackets] with your own details, and share it with your team. It is a starting point, not legal advice: have it reviewed by the right person in your organization.
Do’s and don’ts
Do
- Start with a pilot group and 2 or 3 real use cases
- Set roles, budgets and approval rules first
- Pair your AI champions with colleagues
- Share the Academy, the 30-day plan and the Request library
- Review hours saved monthly
Don’t
- Invite everyone at once with no use case
- Allow every tool and model from day one
- Use levels to rank or evaluate people
- Expect people to figure it out alone
- Look only at spend
The template
AI usage policy: [Organization name]
1. Purpose
We use AI agents to save time and improve our work. This policy explains how we use them safely.
2. What AI agents may do
- Draft, summarize, analyze, research and organize information for our work.
- Prepare messages, documents and reports for a person to review.
- Run routines that a person has set up and approved.
3. What AI agents may not do
- Send, publish, pay, delete or change anything outside our organization without a person’s approval.
- Make final decisions about people, such as recruiting, dismissal or discipline.
- Be used to mislead anyone or to break a law or a contract.
4. Approvals required
- Anything that leaves the organization (emails, posts, payments, changes in other systems) needs a person’s OK.
- A person reviews every AI draft before it is used with a client, a partner or the public.
5. Data we must not share
- [Customer personal information, unless the task needs it and the agent is approved for it]
- [Passwords, keys and access codes: store them in the vault, never in a message]
- [Health, financial or other sensitive records, unless allowed by [the person responsible]]
- [Anything covered by a confidentiality agreement]
6. Private mode for sensitive work
Sensitive work must use an agent added in private AI mode, which uses zero-retention models.
7. Person responsible
[Name, role, email] answers questions, approves exceptions and keeps this policy up to date.
8. Review date
This policy is reviewed every [6 months]. Next review: [date].
Make it yours
- Keep it to one page. If it grows, people stop reading it.
- Name one person responsible, not a team.
- Pair it with the Governance checklist so the rules are also set in Admin.
- A Word version is optional. If you want one, ask us and we will provide it.
Budgets and tools per role
Explain how to set budgets per role and limit which tools members can use in our Crewdle organization, step by step.
Build your weekly delegation list
Look at what I asked my agents in the last month and propose 5 recurring jobs I could hand over completely, with the agent and schedule for each. Then help me set them up one by one.
Replace the [brackets] with your details. The more context you give, the better the first answer.
Frequently asked questions
Is this template legal advice?
No. It is a starting point. Have your own legal or HR contact review it before you roll it out.
How long should the policy be?
One page. Short rules are easier to remember and follow.
Who should be the person responsible?
One named person with enough authority to approve exceptions, often the owner or an admin.
