Two conversations at work
No new ideas. Two scripts. Have them this week.
By the end of this lesson
Walk away with two short conversations to have this week — one with yourself (a paste-rule audit for your role), one with your manager (a 5-minute disclosure check-in). Both pull from lessons 1-4.
No new ideas in this one. Just practice.
Two short conversations you can have this week — one with yourself, one with your manager. Both pull from the four lessons you've just read. The point isn't the exact words. The point is to make this real before the week ends.
Conversation 1 — the paste-rule audit (with yourself)
This is a 15-minute self-check. Open a blank doc. The point is to write down the answers.
- Which AI tool is sanctioned for my role? Microsoft 365 Copilot? ChatGPT Enterprise? An internal Claude? The free public ChatGPT? (If you don't know — that's the first thing to find out, before pasting anything.)
- What did I paste into AI this week? List five recent tasks. Be honest. Customer email? Meeting transcript? Internal financials? A draft press release? Code with credentials in it?
- For each one, run the rule. Would I email this to a stranger outside my company? If yes — public AI was fine. If no — public AI was not fine; was I using the enterprise tool?
- Where I find a mismatch — what do I do now? If a customer's full email got pasted into the public ChatGPT, the data is out. The right move is to tell your manager or your IT team. Quietly raising it early is usually safer than them finding out later.
- What's the rule going forward? Write yourself one sentence. Stick it on a Post-it. Mine looks like: "If I'd email it to a stranger outside the company, I can paste it into the public AI. Otherwise, enterprise tool only."
That's the audit. Fifteen minutes. Now you know where you stand and have a rule that holds.
An Auckland accounts payable lead runs the audit. Four of the five things she pasted this week were fine. The fifth was a supplier invoice with the supplier's full bank details — pasted into the free ChatGPT to ask for help with formatting. Those bank details are now outside her company's boundary. She emails her IT lead the same day, factually: <em>"This happened, here's what was pasted, here's what I'll do differently."</em> No drama. The IT lead is glad she said something. Audit served its purpose.
Conversation 2 — five minutes with your manager
Different shape. Same lesson — disclosure beats hiding. The point is short, factual, friendly. Not an apology. A check-in.
Send a calendar invite. Subject line: "Five minutes — quick AI check-in." Pick a normal one-on-one slot. Don't make it dramatic.
Roughly the script:
"I wanted to do a quick AI check-in — five minutes, nothing serious."
"Here's how I'm using AI for [my role]. I'm using [the company tool] for first drafts, summarising long documents, and comparing options. I'm reading every output before it goes out, and editing what AI got wrong. The judgement on the work is still mine."
"On disclosure — for client work and anything attributed to my judgement, I'm adding a one-line note that AI assisted. For routine internal drafts, I'm not — same as I wouldn't disclose using spell-check. Does that sound right to you?"
"On meetings — for HR conversations, sensitive one-on-ones, anything off-record, I'm switching the AI notetaker off. Should we do anything different as a team there?"
"Anything you'd want me to do differently?"
That's the conversation. Five minutes if your manager is brisk, fifteen if they want to think out loud. Both are fine.
Why this works: it gets ahead of any quiet concern your manager might have. It documents that you're using AI well, not hiding it. It opens the door for them to flag company-specific rules you might have missed.
Why this matters: the most common shape of trust damage in 2024–25 was a manager finding out later that AI was involved in something the employee never mentioned. The five-minute check-in is a small thing that can save a lot of later confusion.
A finance manager in Dublin sends the invite Monday for a Wednesday 10am slot. The conversation runs eleven minutes. Her director thanks her for raising it, mentions a company policy about AI-generated charts that she didn't know existed, and then says "please do this with your three reports as well." The conversation didn't reveal she'd been using AI badly — it revealed she was using it well, and gave her director quiet confidence in how she was handling it.
Try this — pick the dates now
Both conversations only work if they happen. Don't file them under "someday."
Right now, do two things:
- Block 15 minutes in your calendar today or tomorrow for the paste-rule audit.
- Send a calendar invite to your manager — "Five minutes — quick AI check-in." Pick a slot in the next seven days.
That's it. Two artefacts. Two dates in the calendar. The lessons are now real.
Keep this
- Rule — The work-AI conversations are short. The audit is honest with yourself. The check-in is honest with your manager. Both pay off because they happened, not because they were perfect.
- Phrase — "Five minutes — quick AI check-in." — the calendar invite that costs you nothing and protects your reputation.
- Don't — Don't make either conversation more formal than it needs to be. Don't print scripts. Don't read from your phone. Use your own voice and your own examples.
Pop quiz, no marks
-
Which conversation will you have first this week — and when?
Show answer
There's no right answer. The right answer is the one with a date and a time, not "sometime soon." Both are short. Pick one and pick when.
What this practice produces
- The paste-rule audit — A 15-minute self-check that maps your last week's AI use against the safe-paste rule. Identifies any mismatches and the right next step. Synthesises lessons 1 and 2.
- The five-minute manager check-in — A short calendar invite + script for telling your manager how you're using AI, what you're disclosing, and how you handle AI in sensitive meetings. Synthesises lessons 3 and 4.
One last thing — keep these in your head
Twelve cards from the four lessons. No marks. Click to flip. Mark "got it" or "show me again."