After-hours AI answering that books a furnace callback

For a garage door company adding furnace repair and installs, we built an AI voice agent and an AI chat that cover both service lines from one HighLevel account. An unanswered call goes to the AI, and the caller also gets a text back. Whichever way the customer comes in, the AI's work ends at a single booking: a 15-minute slot on a calendar kept for callbacks and nothing else.

That callback is the hand-off. In-home visits stay in the field-service software the business already used, and there is deliberately no connection between that software and HighLevel. The AI never has to guess at a technician's day, and the person who returns the call books the visit where visits have always been booked. In the pipeline, Callback Scheduled sits between Contacted and In-Home Visit Scheduled, so the board shows who is still waiting on a call.

Furnace only. The owner's rule was furnace work and no air conditioning, and it reaches everything the customer can hear or read. The prompts, the knowledge base, the follow-up messages and the reactivation copy speak about furnaces alone. On the website we revised the furnace service page that was already live and added furnace FAQs, so a caller who checked the site first hears the same thing from the AI.

No prices. Some trades get flagged for publishing prices, so no figure is hardcoded anywhere the AI reads, and the AI says a firm quote needs someone to look at the job in person. Install callers who want to think about payment can start on their own: the homepage and the furnace page carry a financing pre-qualify button.

What our QA found. Our test calls on the voice side caught bookings that did not complete and confirmation times read back wrong. That repair is still open as of this page, and we would rather say so than call booking finished.