About half of American workers now say they use AI in their jobs in some way, according to Gallup’s survey data from the spring of 2026. For a lot of them, it started the same way: a coworker mentioned a chatbot that writes emails in seconds, they tried it, and it worked.
That’s fine, as far as it goes. The trouble is what goes into the chatbot. Here’s the scene from my new book that I see play out in real workplaces all the time: a capable, well-meaning credit union teller pastes a member’s name, account digits, and the details of a disputed charge into a free chatbot, so it can “make the letter sound more professional.” The letter comes out better. The member’s private information is now somewhere the credit union can’t see or control.
Nobody was trying to do anything wrong. That’s exactly why it happens so often. So before you paste anything into ChatGPT, Copilot, Gemini, Claude, or any other AI tool at work, check three things.
1. Is this tool approved for work?
Your personal chatbot account isn’t a work tool, however convenient it is. When you use a personal account, your employer can’t see where the information went, whether the account’s settings allow your conversations to be used to train the AI, or what happens to those conversations when you leave the job.
Many employers now provide an approved AI tool, often built into the email or document software you already use. Business versions usually come with different privacy terms from personal accounts, which is a big part of why employers insist on them.
This isn’t a small technicality. In 2023, after an engineer uploaded confidential source code to ChatGPT, Samsung banned staff from using ChatGPT and similar tools and warned that breaking the rules could lead to discipline, including termination. People in IT call unapproved use like this “shadow AI,” and many employers treat it as seriously as emailing confidential files to a personal address.
What to do: search your intranet, employee handbook, and email for “AI,” “generative,” and “acceptable use.” If you can’t find a policy, ask your manager in a short email which tool you’re allowed to use. That email also marks you as someone who’s careful, which is a good reputation to have right now.
2. What’s in what I’m about to paste?
Even in an approved tool, include only what the task actually needs. And in any tool that isn’t approved, never paste:
- Customer, patient, member, or client information: names, account numbers, addresses, dates of birth, medical or financial details, or anything that identifies a real person or describes their situation.
- Employee information: reviews, pay, discipline, medical leave, or complaints about a coworker.
- Passwords, access codes, and security details. Never, in any tool.
- Confidential business information: pricing that isn’t public, contracts, unannounced plans, designs, or code.
Here’s the habit that makes this easy: use stand-ins. Most of what makes a draft useful has nothing to do with the real names and numbers. That teller didn’t need the member’s details to get a better letter. He could have typed:
Write a warm, professional letter from a credit union to a member about an overdraft dispute. Our review found that two of three disputed charges were valid and one was an error, which we’ve refunded. Use [MEMBER NAME] where the name should go.
Same quality of letter. No private information. Fill in the real details yourself, in your own systems, before anything is sent.
3. Have I checked the result?
AI tools can be confidently wrong. They produce smooth, plausible answers, and sometimes those answers include a made-up statistic, a policy your company doesn’t have, or a number that’s simply incorrect. They don’t flag their own mistakes.
At work, that matters, because as far as anyone else is concerned, if your name is on it, you wrote it. Before anything leaves your hands, check every number, date, name, and policy statement against a source you trust, read it the way the recipient will, and ask yourself one question: would I sign my name to every sentence?
The good news is that this checking is exactly where your experience pays off. Anyone can get an AI tool to write fast. The person who knows the work well enough to catch its mistakes is the one an employer can’t do without.
The short version
Approved tool. Nothing confidential, so use stand-ins. Check everything before it goes out. Those three habits let you get the time savings without the risk, and they’re the foundation of the first two chapters of AI Without the Hype: Smart Moves to Protect Your Job and Grow Your Career.
