The shop, before

The customer is a 28-tech HVAC shop covering a four-county service area in a Sunbelt state. The trade mix is roughly 60% residential service-and-repair, 25% residential install, and 15% light commercial. Roughly 45% of annual revenue lands in the four hottest months of the year — the canonical HVAC concentration problem.

Their CSR team — six full-time, two part-time during peak — was sized for normal-day volume. Heatwave volume was a different problem entirely. On the worst days, inbound rate exceeded CSR throughput within the first 90 minutes of the day, and stayed there until close.

  • The capacity wall. On peak days, CSRs were averaging 4–6 calls in queue continuously. Hold times pushed past 8 minutes. Drop-off rates from the queue exceeded 30%.
  • The triage problem. No-cool calls from a household with a child or elderly resident need same-day priority. No-cool calls from a customer planning a system replacement need a sales-track appointment. CSRs in queue-triage mode have no time to ask the questions that distinguish them.
  • The after-hours collapse. The shop did not operate a 24/7 line — it routed after-hours to an answering service that took messages. By morning, the urgency calls had already booked elsewhere.

"Our CSR team was telling me they couldn't actually hear customers anymore. They were just trying to clear the queue. That's not customer service."

— Shop GM (anonymised at customer request)

What we deployed

Standard 14-day deployment timeline. Three channels live:

  • Missed-call text-back on the main shop number — fires within 60 seconds of any unanswered call.
  • Web chat on the homepage and on the no-cool / emergency-service landing pages, with severity-aware greeting copy.
  • Overflow voice routing — calls that wait in the queue past 90 seconds are offered the option to drop off and continue by SMS with the AI Employee.

The HVAC qualifier was tuned around the four highest-leverage data points for this market: system type and age, primary symptom (no cool, no heat, intermittent, weak airflow), severity flags (infants, elderly, asthma, high outdoor temperature), and last-service date. Severity flags routed directly to same-day emergency dispatch logic; everything else went to the next available scheduled slot.

The 90-day comparison

We compared 90 days post-deployment against the matching 90-day window from the prior year. Outdoor temperature data was within 4°F average across the comparison windows, so this isn't seasonal noise.

Before · 90 days (prior year)

  • 3,891 inbound leads captured
  • Peak-day queue drop-off: 30–34%
  • Avg first response peak day: 11 min
  • After-hours: 0 same-night bookings
  • Heatwave-week emergency capture: ~22%

After · 90 days

  • 6,158 inbound leads captured (+58%)
  • Peak-day queue drop-off: 6%
  • Avg first response peak day: 42 sec
  • After-hours: same-night booking by AI Employee
  • Heatwave-week emergency capture: 76% (3.4x)

What the AI Employee actually did

The most measurable behaviour change was during the worst hours of the worst days — the heatwave-week peaks where this shop historically lost the most money.

On a peak heatwave day, the AI Employee handled 180–240 inbound conversations in parallel. Of those:

  • About 35% were severity-flagged emergency calls (vulnerable household, indoor temperature above safe threshold) — routed to same-day dispatch.
  • About 40% were standard service calls — booked into the next available scheduled slot, usually 2–4 days out.
  • About 15% were install / replacement inquiries — qualified and routed to the sales track with full context.
  • The remaining 10% were filtered (out-of-area, warranty referrals, prior-customer follow-up) without burning CSR time.

The CSR team's day-to-day changed shape. Instead of triaging an endless queue, they handled the 15% of conversations the AI Employee escalated — ambiguous severity, customer-relationship moments, complex commercial jobs. Higher-value work, lower-stress days.

"The first heatwave after we went live, I checked our drop-off rate at 4pm expecting the usual carnage. We were at 6%. I made my CSR lead double-check the report."

— Operations director

What this didn't fix

  • It didn't add tech capacity. On the worst day of the comparison window, the shop was still booked-out for emergency same-day slots by 11am — the AI Employee just made sure those slots filled with the right jobs.
  • It didn't reduce CSR headcount. The CSR team's work changed shape; the team didn't shrink.
  • It didn't fix install close rates. Sales conversion is downstream of intake. The AI Employee delivers cleaner, better-qualified leads to the sales track — but closing them is still a sales-team problem.

What it would cost a similar shop

For an HVAC shop of similar size and seasonal concentration, the typical setup is:

  • Setup and 14-day deployment — one-time fee.
  • Monthly subscription — based on inbound volume and channels deployed.
  • Volume-uncapped pricing for high-seasonality customers, since the value lives in the peak-day spikes.

For specifics tuned to your shop's size, volume, and trade, book a 20-minute call.