Around half the clinics and consultancies we speak to are already using ChatGPT for something — drafting emails, summarising notes, sketching out training material. The half that aren't usually say the same thing: we're not sure what's safe.
Here is the short version of what we tell them.
Three rules that cover 90% of the risk
shield The non-negotiables
1. Never paste anything you wouldn't email to a stranger. Client names, NHS numbers, financials, login details — all out.
2. Use a business account, not a personal one. ChatGPT Team and Enterprise don't train on your conversations by default. Free and Plus do.
3. Keep a running list of what works. The prompts that save you an hour are worth more than the AI itself.
The first rule is the one most teams break. The pattern is always the same: someone is in a hurry, the easiest way to summarise the document is to paste the whole thing in, and the prompt gets a name or an email address attached. Once it's pasted, it's gone — you can't pull it back from someone else's model.
What "safe" actually looks like in practice
Safe usage tends to share a few characteristics. The team has a written list of what's in scope (drafts, summaries of your own internal notes, structured outputs from anonymised data) and what isn't (anything containing a third party's information, anything that becomes a record on a client file). They use a paid business plan with data controls enabled. And there's one person who owns the prompt library — so when a prompt works, it gets shared, and when something goes wrong, there's somewhere to put the lesson.
“The biggest mistake we see is treating ChatGPT as a private notebook. It isn't — but it can be a very useful colleague if you give it the same boundaries you'd give a new hire.”
If you only do one thing this week
Pick the single most-repeated piece of writing your team does — appointment confirmations, course intake emails, project briefs — and build one really good prompt for it. Save the prompt somewhere shared. Test it on three real examples. That's the difference between "we use ChatGPT" and "ChatGPT saves us an afternoon a week".
If you'd like a hand auditing what your team is already doing and building that first prompt properly, get in touch. It's usually a one-call conversation.