Three AI rules for your team (before someone pastes client data into ChatGPT)
Short answer: your team is probably already using AI, whether you’ve said yes or not. You don’t need a 20-page policy to fix that. You need three rules everyone can remember: no client or personal data in free tools, AI drafts and a person decides, and you own what it says. Print them, stick them on the wall, and name who approves new tools. Safe use isn’t complicated. It’s three habits.
Your team is already using it
Ask around the office, or the ute. Someone is using ChatGPT to tidy an email. Someone else pasted a customer complaint in to “make it sound nicer”. The office person is using it on quotes.
That’s not a bad thing. It’s people trying to save time. The problem is nobody told them where the lines are.
And people don’t always say. In a 2025 global study by KPMG and the University of Melbourne, 57% of employees said they hide their use of AI and present AI work as their own (KPMG). That’s a global figure, not Australian small business, but it’s a fair warning. If you haven’t set rules, you probably don’t know what’s going in.
Banning it doesn’t work either. A ban most staff quietly ignore is worse than three rules they actually follow.
Rule 1: no client or personal data in free tools
Names, addresses, phone numbers, health details, financial details and anything a client told you in confidence stay out of free AI tools.
Why: free versions of some AI tools may use what you type to train or improve them, depending on the tool and its settings. Once it’s in, you can’t take it back.
What to do instead:
- Use a business account with training switched off. Many businesses already pay for one inside Microsoft 365 or Google Workspace. Check the AI you’re already paying for before you buy anything.
- Take the details out first. “Customer in Penrith, leaking tap, wants it fixed Friday” works just as well as the full name and address. Add the details back in your job software or email, not in the chat.
- Check your own settings. Plans and settings change. Look at the tool’s own page, and the OAIC’s guidance on privacy and commercially available AI products.
If you’re not sure whether something counts as personal data, treat it as if it does.
Rule 2: AI drafts. A person decides.
AI is good at writing words. It isn’t good at knowing what’s true for your business. So it writes the draft, and a person checks it before it goes out.
These always get a person’s check:
- Prices. AI makes numbers up when it doesn’t have yours. (More on that in don’t let ChatGPT price your jobs.)
- Safety. Anything about how a job is done safely, or what a standard says.
- Contract wording. Terms, warranties, guarantees, refunds.
- Anything with your licence or registration on it. If you’d be the one answering for it, you read it first.
A simple test for staff: if this went wrong, would the boss want to have seen it? Then the boss sees it.
Rule 3: you own what it says
If your chatbot, your email or your website says it, you said it. Customers and regulators will treat it that way.
“The AI wrote it” isn’t a defence. If an AI-drafted email promises a Friday finish, the customer expects Friday. If a chatbot on your site quotes a price, the customer expects that price.
So anything that goes out under your logo gets the same care as if you’d written it yourself. That’s why Rule 2 matters.
Who decides which tools are OK
This one sits with the owner. Not “the IT person”, not whoever found the tool first.
Keep it short:
- One list of approved tools, and which account to use for each (the business one, not a personal one).
- One person who approves new tools. Usually you.
- One place to ask. “Not sure? Ask before you paste.”
When someone finds a tool that genuinely saves time, great. They bring it to you, you check the plan and the settings, and it goes on the list.
The one-page rules sheet for the wall
Copy this, put your business name on it, print it and stick it up. Change the wording to suit your team.
[Business name]: how we use AI
1. No client or personal data in free AI tools. Use our business account: [tool name]. Take names, addresses and personal details out first.
2. AI drafts. A person decides. Prices, safety, contract wording and anything with our licence on it are checked by [name or role] before they go out.
3. We own what it says. If it goes out under our name, we said it.
Approved tools: [list]. New tools are approved by [name].
Not sure? Ask before you paste.
Then go through it with the team. Ten minutes at a toolbox talk or a team meeting is enough. Ask what they’re already using. You’ll learn more from that than from any policy template.
For tradies with a team
If it’s you, an office person and a few on the tools, the same three rules work. The big ones for a trade business:
- Customer details stay in your job software, not in a chat window.
- Quotes and prices are checked by you before they go out. Every time.
- Compliance and safety wording is checked against the actual rules, not what AI says the rules are.
If you’re just starting with AI yourself, AI for tradies: the first job to hand over is the place to begin.
Where to go next
Three rules get you out of trouble most of the time. The next step is using AI for something that actually saves hours.
WTF is AI? is six short videos and six cheat sheets for business owners, about half an hour all up. Video 3 is these three rules, with a printable sheet. The rest helps you find the job that eats your week and get it working with the AI you already pay for. A$69 in Australia, US$47 elsewhere.
FAQ
Does a small business need an AI policy? If anyone on your team uses AI for work, yes, but it doesn’t need to be long. Three rules on one page, an approved tools list and one person who approves new tools covers most small businesses. Get proper advice if you handle health, legal or financial information.
Can staff put client information into ChatGPT? Not into free tools. Use a business account with training switched off, take names and personal details out first, and check the tool’s own privacy settings and the OAIC’s guidance. If you’re not sure, leave it out.
Should I just ban AI at work? A ban most people quietly ignore leaves you with no rules at all. Three clear rules and a short list of approved tools are easier to follow and easier to check.
Who should approve AI tools in a small business? The owner, or one person the owner names. Not whoever found the tool first. Keep one list of approved tools and the account to use for each.
What if AI gets something wrong in an email to a customer? You own it, the same as if you’d written it. That’s why a person checks anything with prices, safety, contract wording or your licence on it before it goes out.
General guidance, not legal, privacy or IT security advice. Check your own plans and settings, and get advice for your industry if you handle sensitive client information.