An AI policy your team will actually read

The whole starter policy, in plain language a new hire can absorb in five minutes. It covers what staff may and may not paste into AI tools using a green/yellow/red system (green is fine, yellow means ask first, red never leaves the building), with concrete examples pulled from the data your business actually handles, not vague categories. It names one person who approves new tools and exceptions, sets a no-blame rule for reporting paste mistakes fast, and ends with an acknowledgment line so you know who has read it.

It also produces the rollout email: short, human, and clear that this is a how-to-use-AI policy, not a ban. What it doesn't do is enforce anything. A policy is rules on a page; enforcement is a separate problem.

One session, four steps. First, gather five facts before you open the assistant: which tools are in use today and on what account types, what data would hurt you if it leaked, who approves exceptions, what regulations or client contracts apply, and where the policy will live. Second, run the policy prompt — it locks the structure to seven sections, caps the length at 600 words, and forces the assistant to write [FILL IN] instead of inventing facts you didn't provide.

Third, run the rollout email prompt in the same chat. Fourth, work the QA list: make every red-list example concrete, resolve every placeholder, walk two real scenarios past the policy, and confirm the approver knows they're the approver. If you're in a regulated industry, a lawyer reviews before rollout — the recipe says so plainly, because this is a starting draft, not legal advice.

Honest limit: a policy tells people what not to paste, and that's all it does. It works while AI use means a handful of people drafting emails. Once client data flows through prompts as part of real workflows, across teams and tools, a document can't protect you — the workflows themselves have to be built so sensitive data never touches a consumer chatbot, on approved tools with proper data agreements, with someone accountable for the whole system. That's what we install, and a free audit will show you where your current AI use actually stands.