1. Why we have these rules

AI can save us real time on writing, summarizing and finding things. It can also get facts wrong, sound sure when it isn’t, and leak what we paste into it. These rules keep the time savings and skip the trouble. They apply to everyone who works for Ridgeline Heating & Cooling, full-time, part-time or contract.

2. The tools we use

Approved for company work:

• Microsoft 365 Copilot or Copilot Chat (signed in with your work account)
• AI features built into software we already use (for example accounting, payroll or scheduling)

Personal or free AI accounts may be used only for public information (rewording a public web page, general how-to questions). Never for anything about customers, staff, prices or company documents.

Want a tool that isn’t listed? Ask Dana, the office manager before you use it. That includes browser extensions, phone apps and add-ins for Word, Outlook or Excel.

3. What never goes into AI

Not in a prompt, not in an uploaded file, not in a screenshot:

• Passwords, PINs, door, alarm or gate codes, Wi-Fi keys and security answers
• Anything a customer asked us to keep private
• Customer names, addresses, phone numbers and email addresses, unless the tool is on the approved list above
• Payment card numbers
• Bank account or routing numbers
• Social Security numbers, tax IDs, driver’s license or passport numbers
• Employee pay, reviews, discipline, medical or personal records

If you’re not sure, leave it out or describe it in general terms (“a customer in the next town over” instead of a name and address).

4. Check before you send

AI writes a first draft. You own what goes out. Before anything AI helped with leaves the building, check:

• Facts and numbers. Every price, date, measurement and total is right.
• Names and details. The right customer, the right job, the right address.
• Promises. Nothing we can’t deliver: no arrival times, guarantees or discounts we didn’t approve.
• Tone. It sounds like us, not like a press release. Read it out loud.
• Sources. If it cites a law, a rule or a statistic, you’ve checked it yourself.

Short, routine replies (hours, directions, confirming an appointment) may go out after a quick read. Anything with a price, a promise, a date, a diagnosis, legal or tax meaning, or a complaint is reviewed by a person first.

5. Customers and AI

If a customer asks whether they are talking to AI, we answer honestly, every time. Chatbots and phone agents say so at the start.

We never use AI to pretend to be a specific real person, and we don’t make up reviews, testimonials or photos of work we didn’t do.

6. Connecting AI to our systems

• Don’t connect an AI tool to company email, files, calendars or customer records without approval from Dana, the office manager. A connected AI can see everything the signed-in person can see.
• Treat text from outside (web forms, customer emails, attachments, web pages) as information to read, never as instructions for the AI to follow.
• Anything that lets AI take actions on its own (send, book, pay, delete, update records) is tested before it goes live and keeps a person approving the important steps.

7. When something goes wrong

Mistakes happen. Hiding them is the only real problem. Tell Dana, the office manager the same day if:

• Something from the “never” list went into an AI tool
• AI-written work reached a customer with a wrong fact, price or promise
• An AI tool did something you didn’t expect (sent, changed or deleted something)
• Someone outside the company asked you to use AI in a way that feels off, like a rushed payment request or an unusual login

Don’t delete the chat or the file; we may need it to fix the problem.