Write your own AI policy
From AI Fluency Foundations, module 5. diligence: your ai policy.
Everything you have decided about AI so far lives in your head: which tasks you hand over, which data stays inside, how open you are about it. In your head it is a feeling. On paper it is policy: a line you can hold when things get busy, and show when someone asks.
This free lesson walks you through writing that page. It is the second lesson of the final module of AI Fluency Foundations, and it works on its own. Plan 30 minutes, and have your AI chat open.
Your personal AI policy has four parts. Each part is a short set of concrete rules you can check yourself against on a busy Tuesday afternoon.
- When yes, when no. Which kinds of tasks you do with AI, and which never. Start from your real work: list the recurring jobs of your role and mark for each one what AI does and what stays human. A consultant might land on: proposals (AI drafts the structure, I decide the price and the tone), meeting follow-ups (AI transcribes and summarises, I decide what goes in and what stays out). Then name your never-categories explicitly. Mine include the final sign-off on any client deliverable, pricing, and pressing send on a message.
- With which data. What is allowed in an AI chat, and what never is. Make it concrete enough that you do not have to think in a busy week: "my own drafts, public information and anonymised cases yes; client credentials, signed agreements, personal data and figures shared in confidence no." Add where your AI runs, because that decides how strict you need to be: a tool on your own machine and a cloud chat are different rooms. Rule of thumb: what you would not email to a stranger, you do not paste into a chat.
- How I am open about it. When you report AI use, and how. Substantial text production, analysis that ends up in your advice, anything the recipient trusts to be human work: those you report. A spell check does not need a footnote. Write one disclosure sentence you could say out loud to the recipient: "For this report I used AI for the first draft and the summaries; the analysis and the advice are mine." Then write the rule behind it: for which kinds of work do you always say it, and when is it not needed?
- How I handle mistakes. What you do when an AI error slips through anyway. Who carries the responsibility (you), how you fix it, what you change afterwards. Everyone forgets this part, and it is what makes your policy credible. A short checklist before anything goes out helps: facts against the source, does it sound like me, does it cover what was asked, could it mislead or disadvantage anyone.
Write the policy together with AI: have it ask questions, poke holes, sharpen your wording. But every rule that goes in is yours. It is your line.
Two things make the page stick. Give it a review date, a quarter from now, plus a trigger list: a new AI tool, a client asking about your AI use, a situation the policy does not cover. And when you hit a grey area, ask one question first: would I be comfortable if the client saw exactly how I made this? If the answer is no, change the process.
Your turn
- Open your AI chat and start the sparring conversation, for example: "I am writing my personal AI policy for my work as [role]. It has four parts: when yes/no, with which data, how I am open about it, how I handle mistakes. Ask me two sharp questions per part, one at a time, and then help me phrase it."
- Write out your policy: one page, the four parts, concrete rules. Put a review date at the bottom.
- Paste the finished page back into the chat and ask: "Where do these rules contradict each other, and which rule is too vague to act on?" Fix what it finds, but only where you agree.
- Read it once more as if a client were reading it over your shoulder.
Reflect
Which rule in your policy was the hardest to write down, and why: because it costs you something (convenience, speed), or because you do not fully believe it yet? A policy with one rule that chafes is a real policy. Four smooth rules are a poster.
The rest of the course
This lesson is the second-to-last one of AI Fluency Foundations: five modules, one real project of yours, an AI coach that gives feedback on your work in every module, and a certificate at the end. Start with the free lesson above, then decide.
See the full course