The HVAC quote bot that quoted $400 on a $4,000 job.
Derek, owner of a mid-size HVAC company in Austin serving residential and light commercial in the Cedar Park and Round Rock corridor. Summer demand spikes were creating 24 to 48 hour quote-response gaps. The fix nearly created a worse problem.
Derek runs a mid-size HVAC company in Austin, residential and light commercial, serving the Cedar Park and Round Rock corridor. Summer heat in Central Texas creates peak-season demand spikes where quote requests outpace staff capacity. 24 to 48 hour response gaps were costing him jobs to faster-responding competitors.
The build
An AI chatbot embedded on his website that walked customers through a self-service quoting form: service type, equipment age, approximate square footage, symptoms. Based on submitted parameters, the bot generated a preliminary quote range and offered to book a technician. In testing, it performed accurately across dozens of simulated submissions.
What almost went wrong
Two weeks into live deployment, a customer submitted a form for "replace central air unit" and selected an equipment size two tiers too small. She was guessing based on a sticker she had misread. The bot generated a quote of $387 based on her inputs. She accepted, paid a $150 deposit via the integrated payment link, and posted a positive Google review before Derek's team had reviewed the submission.
Derek honored the deposit as a goodwill credit, redid the quote manually, spent two hours on damage control, and kept the customer. He also absorbed the cost difference and the operational disruption.
The fix
A human-in-the-loop review gate was added for any auto-generated quote exceeding $500 before delivery. The bot now flags those as "pending technician confirmation" and sends: "Your request is in. Confirmed number within two hours." Jobs under $500 still flow automatically. The system now runs cleanly with zero repeat incidents.
The failure was not the AI. It was deploying AI without a risk-calibrated gate. The fix costs almost nothing and eliminates nearly all liability.
What this engagement is, and is not
The Derek engagement is one of the foundational near-misses in the When AI Fails 2026 Report. The pattern that generalized across every quoting and pricing automation we have shipped since: every automated quoting system needs a threshold above which a human reviews before the number leaves the building. Set the threshold where the cost of a wrong quote becomes greater than the cost of a two-hour delay. In HVAC and roofing, that number is usually between $500 and $1,000.
Talk to us about a HVAC build.
Free AI audit first. We read your site, your reviews, your visible stack, then return a prioritized roadmap with a savings estimate. No deck, no upsell. Sometimes we tell you not to build anything yet.
Related reading
- 01
The roofer who trusted his legal pad.
Houston, TX
Adjacent home services. The walkthrough versus the paperwork.
- 02
The law firm chatbot that made up case law.
Boca Raton, FL
Different vertical, same lesson on scope and gates.
- 03
The accountant whose AI sent the wrong tax summary.
Houston, TX
A 15-second human gate prevents a regulatory event.