How a 2-truck HVAC shop booked 28 extra jobs a month in 21 days

The shop runs two trucks out of a small yard on the east side of town — an owner-operator in the field most days, one full-time tech on a swing rotation, and a part-time office manager who answers the phone between noon and four on weekdays. The team is small by choice; the owner is the senior tech on every diagnostic and the office manager covers only what the field cannot. In the six months before the engagement, lead volume was healthy: forty to fifty inbound calls a week, another fifteen web-form submissions, and a steady trickle of referrals that came in as texts to the owner’s cell. What the owner could not reconcile was the booking rate. The trucks had open days on the schedule; the office manager had open hours; the lead count was high; the booked-job count was not.

We pulled a month of records from ServiceTitan, the call log, and the web form submissions before we wrote a single line of workflow. The picture was sharper than the owner expected. Of the inbound calls that reached a real person, only about half resulted in a booked appointment. Of the web-form submissions, fewer than one in three became a booked job. The remaining half of the call gap and two thirds of the web-form gap were not bad leads — they were people who reached a voicemail, got a callback twenty-six hours later, and never picked up. The owner’s read on it was correct: the bottleneck was the speed and structure of the first contact, not the demand.

For an HVAC shop, the speed of the first contact is the booking. A homeowner whose AC died at 10 a.m. is going to call three numbers and book the one who calls back within an hour. A property manager with a rooftop unit down is going to dispatch whoever responds first with the right triage questions — square footage, tonnage, last service date, whether anyone is on site. Every minute between the first ring and a structured response is a minute the shop is ceding the job to the next name on the search result.

The first workflow we turned on was inbound call triage during office hours. When the office line rang and the office manager was on a service window or already on a call, an AI agent picked up, identified the shop, and ran the structured intake an HVAC dispatcher actually needs: address, system type (central AC, heat pump, furnace, mini-split, packaged unit), problem description, system age if known, whether the homeowner is on site, and whether the issue is an active safety concern. The agent wrote the call to ServiceTitan as a lead with a priority flag based on the urgency field, so the office manager came back to a queue she could work in priority order instead of a single voicemail box she had to parse by hand.

The second workflow was web-form triage. Free-estimate requests and emergency-service submissions from the site landed in a shared inbox as raw emails — the office manager was copying the customer’s address and problem into ServiceTitan by hand every morning, usually around forty-five minutes of work. We routed the same submissions through the intake agent, which extracted the same fields the phone workflow produced and wrote the lead directly into ServiceTitan with the matching priority flag. The shared inbox now carries only the submissions the AI was not confident about; in the first week that was four out of sixty-eight.

The third workflow was the after-hours gap. The owner had been reluctant to add an after-hours AI agent because he did not want to dispatch a tech on a lead that was not actually an emergency. The intake workflow made that decision explicit: the agent collects the urgency field, the system type, and whether anyone is on site, and only writes a true emergency directly into the on-call tech’s queue. Everything else sits as a structured lead for the office manager to call back at 8 a.m. The owner’s reaction after the first weekend was that he had stopped getting woken up for jobs that were not emergencies, and the office manager had stopped arriving to a stack of voicemails.

Twenty-one days in, the numbers were clear. Booked jobs for the month, measured against the trailing three-month average from ServiceTitan, were up by twenty-eight. The owner’s own read: the shop did not add a truck, did not add office staff, and did not raise prices. It booked the work it was already paying to advertise for. Average time from first ring to a structured response dropped from roughly one hundred and twelve minutes to under three. The office manager’s morning shifted from parsing voicemails to working a priority-sorted queue; the owner’s evening shifted from a phone glued to his hand to a single on-call dispatch path that the office manager and the AI both trusted.

A side effect we did not plan for was the quality of the diagnostic visit. The tech arriving on site already knew the system type, the system age, the problem description, and whether the homeowner was the decision-maker. The intake was not a replacement for a diagnostic; it was a faster handoff that let the tech spend the first ten minutes on the problem instead of on the basics. The owner estimated that the average on-site visit got fifteen to twenty minutes shorter, which compounded into a real second job per tech per week.

The pattern is familiar. Two-truck HVAC, two-truck plumbing, three-truck electrical — the shape of the answer is the same: a structured intake that runs whether the office manager picks up or not, a CRM write-back the dispatcher actually trusts, and a triage decision the shop can defend on a Monday morning. The stack is whatever the shop is already paying for; the AI sits in front of the first-contact gap and writes the intake the dispatcher would have written by hand. If your operation is running into the same pattern — strong lead volume, weak booking rate, dispatcher buried in voicemail — the right first step is to measure the first-response gap honestly for thirty days. The number will be uncomfortable; the fix is small.