Last week we modeled a 6-truck HVAC shop losing $9,840 every three weeks to missed calls. The leak is real. The next question is what the fix actually looks like — what an AI answering layer does during a call, what the math says at your shop size, and where the spend stops paying for itself.
No vendor crowning. The "which one for my shop" question funnels to the Operator's Audit — that's the only honest place to answer it.
What an AI receptionist actually does, step by step
When a homeowner calls a service-trade shop after-hours (or during a busy run when the office line goes to voicemail), here's what an AI answering layer does in the 30–60 seconds before a human would have called back:
- Greets the caller in a natural voice. Modern voice models (the ones built on the 2025 generation of speech tech) sound like a polite office receptionist who answers in two rings. Not robotic. Not the "press 1 for billing" tree.
- Captures intent in plain English. "Hi, I think my AC isn't cooling — can someone come look at it today?" The AI parses that into: trade = HVAC, urgency = same-day, problem class = cooling failure.
- Asks the qualifying questions a dispatcher would ask. Is the unit blowing air at all? When did it stop cooling? Single-family or multi-unit? Address? These get logged as structured fields, not as a transcript blob.
- Books the appointment — or routes to a human. If the shop has connected its scheduler, the AI offers two windows and confirms one. If the issue requires human triage (gas leak, water main rupture, electrical sparking), the AI flags it as urgent and pages the on-call owner via SMS.
- Texts the caller back with confirmation. Address, time window, technician name if booked, what to expect, what to have ready. Most homeowners get this text before they've finished hanging up.
- Writes the call into the CRM. Housecall Pro, Jobber, ServiceTitan, JobNimbus — the AI logs caller name, number, address, issue summary, urgency tag, booked time, transcript. The next morning the owner opens the dashboard and sees every overnight call as a new lead, not a lost one.
That's the mechanism. None of it is hypothetical — every step is in production at multiple vendors on the public market as of May 2026. What varies between vendors is voice quality, CRM connector depth, call-volume pricing tier, and how the edge cases (callers asking unrelated questions, callers who hang up partway, callers in non-English languages) get handled. Those are the dimensions an Operator's Audit looks at for your specific shop.
The math at a 6-truck shop
In TT#1 we modeled a 6-truck HVAC shop losing $9,840 every three weeks to missed calls. The math was: 50 inbound calls per week × 62% missed (LeadTruffle 2026 research) × 85% never-call-back × average ticket size × close rate. That's the leak.
What does a $25/mo AI answering layer recover from that leak?
Capture rate. A typical answering AI catches 90–95% of inbound calls (the ones that miss are usually voicemail-loop edge cases or carrier failures). Compared to a typical 6-truck shop's organic answer rate of 38%, that's a ~55-percentage-point capture lift.
Book rate on captured calls. Not every captured call books. Industry data on AI receptionists suggests 30–55% of captured-after-hours calls book an appointment when the AI can offer scheduling slots, vs. 15–25% when the AI just takes a message. The lower number applies if the shop hasn't connected its CRM scheduler.
Net additional bookings per week. For our 6-truck shop:
Calls/week: 50
Previously captured: 50 × 38% = 19
Newly captured by AI: 50 × 92% = 46
Additional captures: 46 − 19 = 27 calls/week
Of those, booking rate ~40%: ~11 additional appointments/week
Revenue impact. At a $385 average HVAC service-call ticket (ServiceTitan 2025 Industry Report, residential mix), 11 additional appointments/week is $4,235/week — modeled. Even discounting heavily for the AI not being a magic close-rate machine — say half of those "additional captures" would have called back anyway and half would have gone to a competitor — you're still recovering $2,000+ per week in pipeline that was leaking out.
Cost of the answering layer: $25–$95/month depending on tier and vendor.
The math at this shop size isn't close. It's one-sided.
At what shop size does the math start working?
The $25/mo math is one-sided for a 6-truck shop. Below that the curve bends.
- Solo / 1-truck owner-operator. The owner usually answers the phone in person between jobs. Call volume is ~10–20/week. Missed-call rate is lower (maybe 30%, not 62%) because the owner is the dispatcher. Modeled recovery at $25/mo is ~$300–500/month — still positive, but not life-changing. The bigger lift here is reclaiming the owner's windshield time from phone interruptions, not lead capture.
- 2–3 trucks. Call volume 25–35/week. Missed-call rate climbs into the 50% range because nobody is dedicated to the phone. Modeled recovery: $1,200–$2,000/month. Math is decisively positive.
- 4–6 trucks. Math runs as modeled above. One-sided.
- 7–10 trucks. AI alone may not be enough on the edge cases — emergency triage, multilingual callers, complex billing inquiries. Hybrid models (AI handles routine, human handles flagged escalations) start being worth their premium. Spend per month: $200–$500. Recovery: still 5–10× cost.
- 10+ trucks. Usually big enough to staff a CSR. The math question becomes "AI augmentation vs. hiring a second CSR" rather than "AI vs. nothing." Different article.
Why this isn't a head-to-head test
Last week's Teardown promised a head-to-head AI receptionist test — five vendors, twelve scenarios, ranked recommendation. We changed our minds. The article you just read is the result. Here's why.
We haven't run those tools end-to-end in a real 6-truck shop. We haven't logged 90 days of call data through any of them. To rank them after a hand-built sandbox test would put us in a role we haven't earned — comparison-broker. That's not what Operator Daily is.
What Operator Daily is: a playbook for getting AI to actually transform a service-trade business. The honest version of TT#2 isn't "which vendor wins?" — it's "what does this category of tool actually do during a phone call, what does the math say at your shop size, and at what threshold does the spend start paying for itself?" Those are the questions the sections above answer.
What we deliberately don't do here
We don't pick a vendor and crown them. We haven't run a 90-day pilot through any of these tools in a real shop. To tell you which one wins for your shop based on a sandbox test would be dishonest.
What we can honestly say: the category has at least five real vendors operating in the US service-trade space in May 2026, with public pricing between $25/mo and $300/mo depending on volume and tier. Their feature sets overlap heavily. Their differences (voice quality on tonally-difficult callers, CRM connector depth for Housecall Pro vs. Jobber vs. ServiceTitan, after-hours human-escalation behavior) matter for individual shops but don't translate to a category-wide ranking.
If you want to evaluate the category yourself, the honest path is: pick 2 vendors whose pricing fits your call volume, sign up for trials in the same week, route 20 real calls through each, then keep the one that booked more. That's a 3-week test in your real shop, not a 12-scenario test in a content marketer's sandbox.
The Operator's Audit — where this gets shop-specific
If you want a recommendation tuned to your actual numbers — your call volume, ticket size, CRM, lead mix, current missed-call rate — that's what the Operator's Audit is for.
The first three audits we run go to the first three readers to reply — total, not per reader. After that the price is $297.
A 12-page roadmap built on your shop's actual numbers — sourced industry benchmarks, ranked fixes with costs, a 90-day implementation plan, and a specific AI receptionist recommendation tuned to your CRM, call volume, and lead mix if that's the highest-leverage move for your shop.
Reply with "audit" and your trade. First three replies wins.
Next Tuesday — TT#3
After lead capture works, the next leak is speed-to-quote. Roofers who quote within 4 hours close at 38%; roofers who take 24+ hours close at 11% (industry research). TT#3 walks the photo-to-quote workflow at under $200/mo. Same brand rules: the mechanism, the math, the shop-size threshold — no vendor crowning.