A $4M operator installs its first AI revenue loop.
Day by day, here's a 30-day install on a composite pest-control shop — Maverick Pest Control: $4M a year, 14 trucks, owner-operator Sarah Kim. The intake. The rubric. The first call that put real money on the board. The day-30 numbers. And an honest “what broke.” It's a composite, not a real customer — every number is illustrative, and the math is shown so you can argue with it.
It's an 8-minute read, right here. If the math looks like it could work for you, request a Revenue Leak Audit at the bottom and we'll model it against your actual numbers.
Synthetic use-case. “Maverick Pest Control” and “Sarah Kim” are composites — not a real customer. Every dollar figure is an illustrative range, not a promised outcome. What's real: the install sequence, the intake rubric, the integration steps, the decision logic, the failure modes. Why publish a composite instead of a customer story? Honestly — we're early, and we'd rather show you rigorously modeled work than a glossy testimonial you can't verify. Named-customer case studies will replace this page as they're published. Until then, the AI front desk in this story answers our own phone line; you can call it partway down and check.
The intake — Maverick Pest Control
Sarah Kim runs Maverick Pest Control out of suburban Atlanta. 14 trucks, 22 employees, $4M a year — a $1.4M core of recurring quarterly contracts plus a long tail of one-time call-outs. She bought the business from the founder two years ago. The last six months, it's been flat.
The intake call runs 35 minutes and produces 11 numbers. The six that drive the ranking:
The rubric scores the four levers (Acquisition, Retention, Pricing, Recapture) and ranks each revenue loop by payback × confidence. You don't install all of them at once. You start with the leak that costs the most. Sarah's top three, in order:
- 01
AI Front Desk — Missed-Call Recovery
32% of Maverick's inbound calls land after 5pm and roll to voicemail — about 220 a month. The intake model writes off one caller in seven as gone for good; at Marcus's two-in-three booking rate and the $340 average ticket, that's roughly $85K a year of bookable work walking out the door. Highest-confidence payback on the board — it goes first.
- 02
Multi-Channel AR Recovery
AR sits at 61 days against an industry 38. A modeled $80K–$120K is stuck in the receivables aging. Pull it forward and it could fund the next loop.
- 03
Recurring Revenue Health Monitor
Autopay churn is silent — failed cards, expired services, billing-address moves. 14% on the $1.4M recurring base models to about $196K/yr leaking out.
Rubric output: Install loop #1 — the AI Front Desk — and ship it in 30 days. Layer in #2 on day 31. Re-run the intake at day 90 to score what's shifted.
Install + the first captured call
Days 1–3: stack mapping. Maverick runs ServiceTitan, a Vonage business line, and QuickBooks Online. The AI Front Desk wires into the existing phone number — no port required — and pushes captured leads into ServiceTitan as new opportunities, source tagged.
Days 4–5: training the agent. Marcus, Sarah's dispatcher, sits for 90 minutes and walks the AI through the decision tree he actually uses — emergency vs. non-emergency, recurring vs. first-time, the four quoting tiers, the escalation rule for a German shepherd in the backyard.
Day 6: soft launch, after-hours only. The AI picks up between 5pm and 8am. Live calls during business hours still go to Marcus.
Tuesday night, Day 7 — the $2,220 call. 9:47pm. A new customer found Maverick on Google and called about rats in the attic. Pre-install: voicemail, callback tomorrow, 50/50 they've already booked the next vendor on the list. Post-install: the AI answered, qualified the rodent-exclusion scope ($340 inspection → $1,880 full exclusion), booked the inspection for 8am, texted Sarah the booking, and dropped a contact card in ServiceTitan tagged “after-hours web.” The job closed Thursday: $2,220 booked, plus an $80/mo recurring plan — from one call the office would have slept through.
The AI in this story answers our phone line, today.
Sarah is a composite. The system isn't. The same AI front desk that fields Maverick's after-hours calls in this report picks up our line, 24/7. Call it — ask about rats in an attic, try to stump it, hang up whenever. That's the product doing the job, before you've given us an email address or a dollar.
Call the live front desk: +1 (479) 265-2900A real call, answered by the AI. It qualifies, answers questions, and can book your audit call with Nathan. No human standing by, no capture, no callback unless you ask for one.
Tuning the loop
Eight nights of after-hours capture between the soft launch and the audit: 61 calls answered, 39 booked — a 64% booking rate. On day 14 we sit down with Sarah and Marcus and replay every one of the 61. The wins are obvious. The losses are more useful.
Three changes after the audit:
- 01The bee/wasp differentiator — the AI kept booking wasp calls into a Friday slot, but Marcus only runs the wasp crew Tues/Wed. Fix: train the AI on the real tech-skill calendar, not the generic appointment calendar.
- 02SMS callback for hangups — 6 callers hung up in the first 15 seconds (didn't like the AI voice, or impatient). We added a 2-minute-delay SMS: “Sorry we missed you — text us the address and we'll book the visit.” Recovered 4 of the 6.
- 03Existing-customer handoff — 3 recurring customers called after-hours about a missed quarterly service, and the AI tried to book them as new leads. Fix: look up ServiceTitan customer records before quoting; on a hit, route to the recurring-customer flow (apologize for the miss, reschedule, no new opportunity created).
Day 21: full 24/7 cutover. The AI now handles every inbound call outside business hours, plus daytime overflow when Marcus is on another line. After the audit fixes land, the booking rate climbs from 64% to 71%.
The honest failures
No 30-day report is honest without this section. Five things that went wrong — and what we did about each:
The German shepherd in the backyard
The AI booked an inspection without flagging the dog. The tech showed up, couldn't reach the crawl space, drove away. Cost: $85 of unbillable truck time. Fix: a dog/livestock/access-issue qualifier, now a required field on any exterior inspection.
The voicemail-greeting trap
Day 9. Maverick forwarded an overflow line wrong, and the AI picked up Marcus's personal cell — then introduced itself as Maverick's front desk. Marcus's golf buddy was confused. Fix: an ‘is this a Maverick-owned line?’ check on phone-line provisioning.
ServiceTitan tag collisions
Day 12. The AI created duplicate opportunity records for one caller — two calls, two records, one customer. ServiceTitan dedupe missed them because the AI used slightly different phone-number formats. Fix: enforce E.164 normalization on the AI side.
The accent the AI couldn't follow
Day 16. An elderly caller with a strong rural Georgia accent tried to book a hornet removal. The AI asked for clarification four times, then transcribed the address wrong. The caller hung up. Fix: a confidence-threshold fallback that routes ambiguous calls to a callback queue instead of pushing through.
The $0 quote
Day 22. The AI quoted “$0” for a service the customer wasn't eligible for under the recurring plan. They thought it was a free visit. Fix: a hard quote-floor of $89 (minimum trip charge) plus a script for explaining recurring-plan exclusions instead of returning zero.
None of these were catastrophic. All of them were tuneable. The install isn't “set and forget” — weeks 1–3 are real tuning work, which is why we own that window on a managed install.
Business impact — the dollar number
30 days after the AI Front Desk went live — the numbers from ServiceTitan and the call logs:
What we'd actually claim: Not the $47K — most of those callers would have rung back at 8am and booked anyway. The model only credits the slice the intake wrote off as gone for good, which nets out to $40K–$64K a year of incremental after-hours revenue — 24/7 phone coverage without adding headcount. The AR and Recurring Revenue loops start in month 2, modeled at another $90K–$140K over the following 90 days. Payback on the install itself depends on your scope; it's the first number we model on your audit call. Illustrative ranges on a composite operator — not a promise.
Let's model the math on your actual numbers.
A thirty-minute audit call. Bring your call logs, your lead sources, your last 90 days of revenue. We'll run the same rubric we ran on Maverick and tell you what the loop could pay back — even if the answer is “not yet.”
Day 30, in the operator's voice
Sarah is a composite — so read this for what it is: the verdict this install is designed to earn by day 30, said the way an operator would say it:
“The Tuesday-night rat call is when I stopped being skeptical — $2,220 booked while every one of us was asleep. Marcus can actually go home at 5 and the phone doesn't follow him. I don't wake up to six voicemails to triage before the trucks roll. The week-2 audit mattered more than the launch — it's not magic, it's a system that had to be tuned to how we actually run. Now that it's tuned, it just does the work. Next loop starts next month.”
Next 30 days: install the Multi-Channel AR Recovery loop and wire it into QuickBooks Online + Maverick's Twilio account. Re-run the intake at day 90 to score what's shifted.
What's synthetic, what's real.
- ✓Real: the install sequence, the rubric, the integration map (ServiceTitan + Vonage + QBO), the audit cadence, the failure modes, the AI tuning approach.
- ✓Real: the dollar ranges are calibrated to operators we've modeled in this size band ($3M–$6M home services with recurring revenue).
- ✗Synthetic: “Maverick Pest Control,” “Sarah Kim,” the specific calls, the exact dollar amounts. Composite, not customer.
- ✗Not a guarantee: your numbers will differ. The same install on a $1M operator pays back slower; on a $10M operator, faster.
Ready to find out what this looks like on your numbers?
Book a Free Revenue Leak Audit. We bring the rubric. You bring the call logs and lead sources. You leave with three ranked install candidates and a straight answer on whether the math pencils. No deck, no pitch, no follow-up sequence.
Honest go / no-go at the end. We'll tell you if your vertical or revenue stage isn't a fit for the install instead of selling you something that won't work.