This isn’t theory. This is a client report.
A Windsor HVAC company, 14 techs, 3 office staff, implemented three AI workflows last month. Here’s what actually happened.
Job 1: Call triage bot
The goal: Pick up after-hours calls, capture lead info, book the actual-emergency calls into the on-call rotation.
What worked: Capture rate went from ~40% (voicemail) to ~85% (bot took the info). Two emergencies got properly routed.
What flopped: The bot misread one caller’s accent and booked the wrong address. Tech drove 40 minutes the wrong way. $180 wasted.
Job 2: Estimate write-up
The goal: Tech dictates from the truck, AI cleans it into a customer-ready estimate doc.
Result: Works. Estimate turnaround went from next-day to same-day. Closed-won rate up slightly.
Job 3: Service history summaries
The goal: Before a maintenance visit, hand the tech a one-pager on the customer’s last three visits.
Result: Techs loved it. Customers noticed. One asked if they’d hired a “customer intelligence person.”
The Monday test
Don’t try all three. Pick the one where a bad outcome costs you the least, and run it for 30 days. Measure. Then decide.