For teams that actually use AI at work

If your team is already using tools like ChatGPT, Claude, or Perplexity, you need some kind of AI usage policy. Not a 30-page legal doc. Just a shared understanding of what’s okay, what’s not, and how to handle the grey areas.

This guide is for you if you’re thinking:

“We should probably write something down…”
“I don’t want to over-police this – but also don’t want chaos.”
“Can we just get a clear, lightweight version to start with?”

Yes, we can. It's easier than you think.

1. What AI tools can we use?

Start by being clear about what’s allowed.

Approved tools: ChatGPT (web and API), Claude, Perplexity, GitHub Copilot
Off limits: Tools that retain data for training without clarity, or that generate unreliable or unsafe outputs

Trying something new?

Ask [person/team] before using a new tool. We’ll check privacy, usefulness, and risk.

2. What kind of data is okay to use?

Set boundaries to protect sensitive info:

Don’t share anything personal, confidential, or protected (e.g. names, email addresses, customer data, financial info)

Okay to use:

Public info
Anonymised notes
Drafts of your own writing or code

"Don’t put personal or sensitive info into AI tools unless they’re approved for that use."

3. What are we using AI for?

Help the team understand what it is useful for, and where to be careful.

✅ Common OK uses:

Marketing: blog outlines, social copy drafts, summarising interviews
Coding: debugging, writing tests, boilerplate code
Ops: writing SOPs, summarising meeting notes
Docs: drafting help guides, FAQs, onboarding steps

🚫 Not OK or use with caution:

Writing policies or legal text without human review
Generating public content without checking tone, facts, and spelling
Making hiring, HR or financial decisions
Using unvetted tools to process private data

4. Who’s responsible for checking the output?

Always make it clear that:

AI helps generate ideas, but people are responsible for what gets used.

That means checking for:

Factual mistakes
Misleading tone or bias
Missing context or nuance

Don’t assume the tool got it right. Even when it sounds confident.

5. How do we keep this policy useful?

Reviewed every 3-6 months
Suggestions welcome
[Team/person] keeps track of tool changes and updates guidance

Want a starter template?

We’ve put together a simple, editable AI Usage Policy Template for teams like yours. Not a legal doc, a real world policy.

Check out our template or get in touch if you want help adapting it.

What's An AI Usage Policy? And Why Your Team Needs One

Want More Resources Like This?