🤫 Use AI At Work Without Making It Weird

Every workplace right now has two kinds of AI users. There’s the one hiding it like a hip flask - alt-tabbing away when someone walks past, quietly terrified that using Claude to draft an email counts as cheating. And there’s the one who mentions it in every meeting, forwards AI-written everything, and has somehow made “I asked ChatGPT” their whole personality.

Both of them are making it weird. And honestly, it’s not their fault - this all arrived faster than the etiquette did. There’s no chapter in the staff handbook, no unspoken rule set that everyone absorbed at their first job. So here’s the etiquette guide nobody wrote, built around one idea: keep it boring, keep it honest. Use AI for the unglamorous middle of your work, be straightforward if it comes up, and don’t make it a thing in either direction.

The Quiet Colleague Rules

Think of AI as a very capable colleague who sits next to you but doesn’t attend meetings. Three rules govern how you work with them.

1. Keep it boring. What it is: use AI for the parts of your job nobody would fight over - first drafts, summaries, prep, tidying, reformatting. The middle of the work, not the judgement at either end. You decide what needs doing and you decide what’s good enough to send; AI does the slog in between. How to do it: next time a task lands, split it in your head into “the thinking” and “the typing”, and hand over only the typing. A status update is your knowledge plus twenty minutes of formatting - keep the knowledge, delegate the formatting. The non-obvious bit: the closer a task sits to your actual judgement - a performance review, a decision about a person, anything where you are the point - the less of it you should delegate. Not because you’d get caught, but because those are the tasks your reputation is actually made of.

2. Keep it clean. What it is: watching what goes into the chat box, because that’s where the real workplace risk lives - not in being seen using AI, but in pasting something confidential into it. How to do it: before you paste, strip or swap anything identifying - names become “the client”, figures become “[X]”, and the story still works fine for drafting purposes. If your company has an AI policy, read it (it’s probably one page). If it has an approved tool, use that one for work things. The non-obvious bit: names aren’t the only identifiers. “Our biggest client in Leeds” plus “the December contract issue” narrows things down fast, and combinations reveal what single details don’t. The full rules on what should never go into a chat box live in The Never-Paste List - if you use AI at work at all, that one’s worth five minutes.

3. Keep it honest. What it is: knowing when to mention AI helped, and when nobody needs a press release. The dividing line: mention it when someone asks (obviously, and casually - lying about it is the only genuinely career-damaging move on this page), and mention it when the work is being valued as your personal expertise - analysis you’re presenting as your professional judgement, anything your name is vouching for in a way that matters. Don’t bother announcing it for the everyday stuff, for the same reason you’ve never announced spellcheck, a calculator, or googling something before a meeting. How to do it: when it fits, one breezy mention does the job - “I got a first draft out of Claude and then knocked it into shape”. The non-obvious bit: said casually and occasionally, that sentence actually builds your reputation. It says you’re current, efficient and unbothered… which beats both the hip flask and the press release.

Ten things that are absolutely fine

For calibration - all of these sit comfortably inside the rules, anonymised where needed:

1. Meeting prep - two-minute briefing before you walk in.
2. Minutes into actions - messy notes in, tidy action list out.
3. Status updates - your bullet points, properly dressed.
4. Tone-checking a tricky email - “does this sound passive-aggressive?” (It did.)
5. Handover notes - before leave, so nobody rings you on a beach.
6. First drafts of anything - proposals, adverts, rotas, agendas.
7. Translating jargon - “explain this IT email in plain English.”
8. Proofreading - typos, clunky sentences, that one comma.
9. Rehearsing a difficult conversation - practise the pushback before the real thing.
10. Turning a rant into a professional reply - type what you want to say, ask for what you should say.

Notice the pattern: every single one is boring. That’s the point. The boring uses are where the hours actually go.

The work-safe starter collection

Four prompts covering the four most common “I wish this was already done” moments of a working week. All designed to run on your input - no confidential pasting required.

Nick this one - meeting prep:

I've got a meeting about [TOPIC] with [roles, not names - e.g. "my manager and someone from finance"]. My goal: [WHAT YOU WANT OUT OF IT]. Give me: a 5-line brief on the topic, 3 sharp questions to ask, and the most likely pushback with a calm answer to each. Two minutes to read, max.

Nick this one - minutes into actions:

Here are my rough notes from a meeting (names swapped for roles): [PASTE]. Pull out: 1) every action with who and by when, flagging any with no owner or date, 2) decisions made, 3) anything left unresolved. Don't add anything that isn't in the notes. Then draft a short, friendly recap message I can send to the group.

Nick this one - status update:

Turn these bullets into a status update for [my manager / the team]: [WHAT'S DONE, WHAT'S IN PROGRESS, WHAT'S STUCK, WHAT I NEED]. Format: 3 short sections - Done, Doing, Need. Confident but honest about the stuck bits, no waffle, under 150 words.

Nick this one - handover notes:

I'm off [DATES] and need handover notes. Interview me: ask one question at a time about my live tasks, deadlines while I'm away, who covers what, and where things live. Then build a handover doc so clear that nobody needs to contact me. Include a "if X happens, do Y" section.

These all work in ChatGPT too - the etiquette doesn’t care which tool you use.

A worked example, so you can see it done cleanly

Say you work in the office of a plumbing supplies firm, and a big customer has sent a stroppy email about a late delivery. The weird version: paste the whole email - customer name, account number, order value - into a chat and ask for a reply. The clean version: “A key customer is unhappy about a delivery that arrived four days late due to a courier issue. They’re hinting they’ll take their business elsewhere. Draft a reply that apologises properly without grovelling, explains without excusing, and offers a gesture of goodwill. Firm on the facts, warm in the tone.” Same quality of draft, nothing sensitive shared, and the judgement calls - which gesture, how warm - stay exactly where they belong. With you.

Where this stops working

If your workplace has actually banned AI tools, this guide is not a licence to sneak - that’s a conversation to have (or a policy worth questioning), not a rule to quietly break. If you work with regulated data - medical, legal, financial, safeguarding - your industry’s rules beat anything written here, every time. And no etiquette survives bad work: if AI drafts it and you send it unread, the problem isn’t that you used AI, it’s that you didn’t do your job. The reading is the job.

The five-minute version of this guide

Use the minutes prompt after your next meeting. It’s invisible, it’s five minutes, it makes you look organised, and there is nothing weird about it whatsoever. That’s the energy.

Once the etiquette feels natural, the next level is using it to quietly outshine everyone before the meeting even starts: Be The Most On-It Person In The Meeting.