Team AI Usage Memo: 9 Rules for Safe and Effective Prompting

Rule 1. Protect the data. Before you paste anything. ☑ Ask: "Would I be fine seeing this in the company-wide channel, with my name on it?" ☑ No? Swap names for roles. "Sarah at Acme" becomes "the client." ☑ Never click thumbs up/down on a work chat. ☑ Never click "Share" on a work chat. Never click the 'thumb' button; it sends that chat to train the model. Rule 2. Open every link. ☑ Every number, every name, every source. ☑ You shouldn't be the person responsible for presenting fake data to the team and to clients, just because you were lazy enough to double-check the source. Rule 3. Use your brain first. Then prompt: ☑ Write your own 3 options before opening any AI. ☑ Maybe after then: "What did I miss? Compare all of them for our situation." ☑ Never end a prompt with "right?" or "maybe?" Rule 4. Claude won't be accountable. You will be. ☑ If you send anything, you wrote it. ☑ Can't explain a sentence yourself? Not worth sharing. ☑ Raw AI idea? Better to label it: "Claude suggested this, worth exploring?" Unedited AI output turns your work into my work. I'm not your proofreader. Don't make my life difficult. Rule 5. Go to brain gym. ☑ Once a day, write one first draft yourself. ☑ Do one calculation by hand. ☑ Make one decision before asking. It's not that hard. You used to do it before ChatGPT 3 years ago. Don't lose your intelligence to AI for short-term laziness and efficiency. Rule 6. Don't connect everything. ☑ A connector can read your whole inbox, your whole Drive, your whole Slack. Most of them you'll never use. ☑ Connect only what it worth, read-only when possible. ☑ Never plug a work account into a personal AI account. Review what's connected once a month. If you are still unsure? Ask me first. Rule 7. Treat Claude like someone who joined yesterday. ☑ It doesn't know our clients, our jargon, or what happened in Tuesday's meeting. Tell it. ☑ Send the screenshot instead of retyping the numbers. ☑ Tell it to look around first: "Before you start, open every email, doc, and tab that could be relevant." Claude feels dumb halfway through? Open a new chat. Every time. Rule 8. Shorter wins. ☑ You write once. The team reads 10 times. If a 2-line prompt made a 2-page doc, send the 2 lines instead. ☑ Do not use AI for everything. You are smart enough to do the micro tasks faster than AI. Nobody should receive these AI patterns: "Here's the thing," "It's not X, it's Y," "delve," "seamless," a pile of em dashes, a pep-talk ending. Rule 9. Ask once. Ask right. ☑ Give the goal, the context, and what "done" looks like. ☑ Delete "think step by step." It already thinks.