Your staff is already using AI, whether you’ve approved it or not. Some of them are pasting client contracts into ChatGPT to draft a summary faster. Others haven’t touched it at all, because they’re afraid of looking replaceable or afraid of breaking something they don’t understand. I’ve sat across from both types of employees, often at the same firm, and both groups have the exact same problem. Nobody ever explained what the tool is.
I’ve been working in technology long enough to remember when automation meant macros in Excel. Now AI shows up in customer service chatbots, document drafting, data analysis, and half the software your firm already pays for, often without anyone announcing it. That speed is real. So is the risk, especially for people who treat AI like a shortcut instead of a tool with limits.
Why Employees Are Using AI Incorrectly at Work
Most employees using AI incorrectly at work aren’t being careless. They’re being efficient, which is exactly the problem. An associate under deadline pressure isn’t thinking about data handling policy. He’s thinking about getting a first draft done before five o’clock, and ChatGPT does that faster than he can.
I watched this happen at an accounting firm I work with. A staff accountant pasted a client’s full financial statement into a public AI tool to get help summarizing it for a partner meeting. She wasn’t being reckless. Nobody had ever told her that data leaves the building the moment it’s typed into a tool like that, and that it may be used to train a model she’ll never see or control.
AI isn’t intelligent in the human sense. It’s powerful, but it only works as well as the data behind it and the judgment of the person driving it. When that person doesn’t know where the line is, the mistake isn’t malicious. It’s just uninformed.
Is It Safe to Use ChatGPT With Client Information?
No, not with the free, consumer version of ChatGPT or any similar public tool. Anything typed into a standard AI chatbot can be stored, reviewed, or used to train future models, depending on the platform and its settings. For a law office, an accounting practice, or a property management company handling tenant records, that’s client data leaving your control the moment someone hits enter.
There are business-grade AI tools built with data protection and confidentiality in mind, and firms handling sensitive client information should be using those instead of the free public versions their staff download on their own. The fix isn’t telling employees to stop using AI. Most of them won’t, and pretending otherwise doesn’t solve anything. The fix is giving them an approved tool that does the same job without the exposure and a policy of proper usage.
The Other Half of the Problem: AI Fear in the Office
AI fear in the office rarely gets talked about directly, because admitting you’re afraid of a piece of software isn’t something most people want to say out loud in a staff meeting. I’ve talked to paralegals convinced that AI is coming for their job, and office managers who avoid every new AI feature in their software because they’re worried they’ll break something they can’t fix.
That fear isn’t irrational. It comes from the same root cause as the misuse. Nobody sat down and explained what the tool does, what it doesn’t do, and where a person’s judgment still matters more than the output on the screen. Silence gets filled with either overconfidence or avoidance, and I’ve watched both play out in the same office within the same month.
One office manager told me she’d rather retype a document by hand than risk using the AI feature built into her firm’s software, because she didn’t want to be the one who “broke something.” She wasn’t wrong to be cautious. She was never given a chance to be curious instead, because nobody at the firm had made it safe to ask a basic question about a tool everyone assumed she should already understand.
What Fixes Both Problems
A one-page policy memo won’t fix this, and neither will a lunch-and-learn nobody remembers a week later. What works is sitting down with the people using these tools and having a real conversation about what AI is good at, what it gets wrong, and where a person still has to check the work before it goes to a client.
AI should support someone’s judgment, not replace it. That means teaching your team to verify anything AI produces before it leaves the building, whether that’s a draft email, a data summary, or an answer to a client question. It also means being honest that some of them are already using free AI tools with sensitive data, and treating that as a gap to close rather than a mistake to punish.
If you haven’t asked your staff directly which AI tools they’re already using and what they’re putting into them, that’s the conversation to have this month, before you write another policy nobody reads.
Download an example of how to create an AI policy here.
Primary keyword: employees using AI incorrectly at work
Secondary keywords: AI mistakes in the workplace, AI fear in the office
Meta description (154 chars): Half your team is scared of AI. The other half is quietly misusing it. Both problems come from the same place, and neither gets fixed by a memo.
