Your office staff are probably already typing customer names, addresses, and job details into ChatGPT or a similar tool right now. That isn't automatically a disaster, but the real problem is that nobody in the business has said what is okay to type in and what isn't. Defining that boundary is the first thing you need to fix.
Is your office already using ChatGPT with customer information?
A 2026 survey by Wakefield Research for PagerDuty asked 1,250 office professionals at companies with $500 million or more in annual revenue about their AI habits. 66% said they used an AI tool at work even though they believed it was not permitted under company policy. 34% admitted they had entered customer data or information into a public AI tool such as ChatGPT, Claude or Gemini. 43% said they had pasted work-related correspondence into one of those tools.
Those numbers come from large enterprises, but the pattern is not limited to Fortune 500 firms. Small home-service shops often lack an IT department, formal policy documents, and the same level of oversight that big companies have. In a typical three-person office, a technician may copy a service note into ChatGPT to draft a follow-up email, while the shop owner assumes the interaction is harmless because it "just feels like a quick shortcut." The reality is that every time data is typed into a public window, the business loses visibility over where that text ends up.
What's actually risky about it?
When a customer's name, phone number, address, or job details appear in a public chat window, the business no longer controls that information. You cannot see which servers store the text, who else might have access to it behind the scenes, or whether the tool will use those words to train its models later. Even when a vendor says data is not stored, the language is often vague and you are left without proof.
That loss of control creates several practical problems:
- A support agent cannot retrieve the original note if the AI service decides to delete old sessions.
- Future model updates may incorporate your customers' details, meaning the information could appear in unrelated outputs for other users.
- Your team cannot audit who typed what and when, making it impossible to trace mistakes back to their source.
The safest approach is to keep any customer-specific information inside systems that you own or that are explicitly designed for secure handling of service data.
What do you actually do about it today?
Pick one trusted person, often the office manager or lead technician, to be the single gatekeeper. That individual decides, on a case-by-case basis, whether a piece of text can be entered into an AI tool or must stay inside your own software. No forms, no extra training sessions, just a clear "yes" or "no" from one accountable voice.
What this looks like done for you
When you work with Key's Touch, we start by mapping every place customer information currently flows through your shop. We look at phone logs, email threads, field-service apps, and any ad-hoc notes that technicians keep on their phones. From that map we replace risky habits with an approved automation built around a single named approver.
You'll see exactly how the process works in practice on our home services page. It lays out what Key's Touch builds for home-service businesses, including intake, follow-up, and support automations, all built around a named approver who signs off on what is allowed.
To help you get started, download the free AI Readiness Checklist. The checklist walks you through five key areas, process clarity, data and tools, team ownership, cost, and security, in just ten minutes, so you can see where you stand before we talk.
Sources (verified live Aug 27, 2026):
- PagerDuty Shadow AI Survey 2026, conducted by Wakefield Research among 1,250 office professionals at companies with $500 million or more in annual revenue: PagerDuty, "Shadow AI Is Happening Within Your Organization".