Doing it safely.

This is the lesson that keeps you out of the stories you've read about. None of it is complicated, and none of it should put you off. Get three things right with ChatGPT and you can use it with confidence rather than worry.

1. Mind what you paste in

The simplest rule, and the one to share with everyone: don't paste confidential or customer data into the free consumer version of ChatGPT. Customer names and contact details, anything covered by privacy, contracts, account numbers, commercially sensitive figures, keep them out of the free tool. On the consumer tier you don't fully control where that text goes, and it may be used to improve the system, so check the data controls.

The good news is you rarely need to. For most jobs you can work with a made-up example or strip out the identifying details and still get a great draft. "Write a reminder for a customer whose invoice is 14 days overdue" works perfectly without a single real name.

2. Use a business setup, and check the settings

When you do need to work with real business information, use a business or Team setup rather than a personal free account. On the Team and business plans your content stays inside your workspace and is not used to train the model, which is exactly what you want. It's the difference between a public counter and your own back office.

Whichever plan you're on, take two minutes to check the data controls in settings. There's a setting that governs whether your chats can be used to improve the model; on a business account you'll want that off, and you should confirm it rather than assume. The defaults aren't always the safest choice, and a quick look keeps your business content yours. We walk through this in plain English in is your business data safe with AI.

3. It can be confidently wrong, so a person checks

You've heard this throughout the course, and here's where it really counts. ChatGPT can hand you a polished, confident answer with a specific figure or fact in it that is simply made up. Confidence is not proof. So a person checks the facts, and a person signs off anything that touches a customer, money or the law before it goes anywhere. Use it to draft and to understand, never as the final authority on something that matters.

The one-page team rule

If anyone besides you uses ChatGPT, write a single page so nobody has to guess. It needn't be formal. Three lines do the job:

• What's fine to use it for: drafting, summarising, rewording, first drafts, with made-up or de-identified examples.
• What never goes into a public tool: real customer data, anything confidential or sensitive. Use the approved business setup for those.
• Who checks and signs off: the human-eyes rule for anything customer-facing, factual, legal or financial before it's sent.

That single page turns AI from a quiet worry into something your team can use with confidence. It's the same governance the broader AI course recommends, kept specific to ChatGPT.