Roll Out a Team AI Policy in One Afternoon using Claude

Why this matters

Your team is already using AI, with or without rules. Someone tried ChatGPT on a proposal, someone used a free tool to tidy a client spreadsheet, and nobody ever said what is okay.

The result is predictable: client details in personal accounts, AI drafts sent to clients unchecked, and mistakes nobody reports because they are afraid of the reaction. Writing the rules takes an afternoon. One leak of client financials costs far more.

In this module you make the decisions and Claude does the writing: a one-page AI policy in plain words, an announcement email from the boss, and a five-question quiz so people actually read it.

What you’ll build

For Northfield Consulting Group, a made-up 18-person firm, you send Claude one request with seven rules: the approved tools, what never goes in, what is fine to use, a person checks every output, be open with clients, who to ask, and a same-day, no-blame rule for mistakes.

Claude writes a one-page policy (about 300 words) in a document beside the chat. You check the three sections that matter most, then get the announcement email and a true or false quiz that tests the rules people get wrong. The text is what Claude really wrote for this request.

What you need

Nothing to start the practiceClaude is a practice copy. No account is needed and nothing is saved or sent.
A Claude account, to do it for realclaude.ai. The free plan handles a request like this, within its usage limits. Use your company account if you have one.
Your own rules, decided firstWhich tools are approved, what must never be pasted, who checks AI work, and one named person to ask. Write them as a short list. The prompts below show the format.
Where the policy will liveA shared drive folder or intranet page your team already uses. You paste that link into the announcement email.

Concepts

3 ideas you’ll use in the practice.

You set the rules, Claude writes them

Ask "write us an AI policy" with no rules and you get a generic one that fits nobody. The decisions are yours: tools, red lines, review, who to ask. Claude’s job is clear, consistent wording.

The request also says "Do not name or quote any laws or regulations." Rules differ by state, country and industry, and a made-up citation in a policy is worse than none. Add those parts yourself with someone qualified.

Check the three sections that matter

Read every line, but look hardest at three: the never-paste list, the rule that a person checks every output before it reaches a client, and the mistake rule.

The mistake rule is the one firms forget. If people fear blame, a wrong paste stays hidden. "Tell Dana the same day, no one will be blamed" means you hear about it while you can still fix it.

An announcement and a quiz turn a document into a habit

A policy nobody reads protects nobody. A short email from the managing partner, with a read-by date, gets it opened. A five-question quiz in a team meeting gets it remembered.

Good quiz questions test the mistakes people really make, like "Can I use my personal account if I delete the chat afterwards?" (False.)

Practice

Do every step yourself in a practice copy of the real tool. It checks your work, and “Do it for me” finishes any step you get stuck on.

ClaudeRoll Out a Team AI Policy in One Afternoon using Claude13 steps · ~10 min · nothing touches your real accounts

Start the practice →Prefer to read? Step-by-step guide

Do it for real

Now repeat it in your own account. Tick each step as you go.

Write down your rules: approved tools and accounts, the never-paste list, what is fine to use, who checks AI work, how you answer clients who ask, who owns the policy, and what to do after a mistake.
Open claude.ai and paste the policy prompt below with your firm, tools, rules and names swapped in. Keep "under 450 words" and "Do not name or quote any laws or regulations."
Read the policy. Check the never-paste list, the human-review rule and the mistake rule word by word. Ask Claude for any change by naming the section.
In the same chat, paste the rollout prompt to get the announcement email and the five-question quiz.
Save the policy where your team will find it, and replace Claude’s link in the email with that link.
Send the email with a read-by date, run the quiz at your next team meeting, and put a review date six months out in your calendar.
If your field is regulated, have a lawyer or compliance person review it before you rely on it.

Downloads

The two policy prompts (.txt)The exact policy request and rollout request from the lesson. Swap in your firm, rules and names.
A finished one-page policy to compare (.txt)The policy from the ChatGPT data controls module, written for a made-up bookkeeping firm.

Quick check

3 questions to check you’ve got it.

1. What should you give Claude before it writes the policy?

Nothing, it knows what a good policy says
Your own rules: tools, what never goes in, who checks, who to ask
A list of every law it should quote

Show the answer

Your own rules: tools, what never goes in, who checks, who to ask. Claude writes; you decide. Without your rules you get a generic policy that fits nobody.

2. Why does the policy say mistakes are reported the same day, with no blame?

So people hide fewer mistakes and you can fix them quickly
Because mistakes do not matter
It is required by law

Show the answer

So people hide fewer mistakes and you can fix them quickly. Fear of blame keeps a wrong paste hidden. Quick, safe reporting lets you limit the harm.

3. Why ask for a quiz as well as the policy?

To make the policy longer
So people read it and remember the rules they are most likely to get wrong
To replace the policy

Show the answer

So people read it and remember the rules they are most likely to get wrong. The email gets it opened; the quiz gets it remembered.

Up next · Module 4

Prompt Writing 101

Your team has the rules. Now teach them to get answers worth checking: Prompt Writing 101 shows why one vague line gets a generic answer, and the four-part prompt that fixes it.

Start module 4 →