Industries · Home services & trades
The phone bill says 400 calls. The board booked 180.
Corveonic builds the office layer for HVAC, plumbing and electrical shops: PO matching against the supply-house statement, warranty claims chased to credit-applied, estimate follow-up inside the rules, and one page on Monday that reconciles to your QuickBooks P&L. A person stays in every loop.
5:40am to 11:00pm
A day on your board
An eleven-truck HVAC and plumbing shop in mid-July. Nothing below is invented; it is how the day is documented to run.
- 5:40am
The on-call tech texts a cracked-heat-exchanger photo from an 11pm job into a group text named SERVICE with nine people in it. The owner sees it at a stoplight. The owner's personal cell is production infrastructure: escalation path, approval workflow, photo archive, after-hours dispatch queue.
- 6:30am
Nobody says the scheduling module. The dispatcher drags from the job tray onto a tech's row, checks tags and invoice subtotal before moving anything, and hits Sync by hand, because another dispatcher may have moved something. Two dispatchers working one board is a race condition.
- 7:00am
The huddle. Truck stock was checked in theory. The lead tech is missing a TXV he needed yesterday and will hit the supply-house counter at 7:30, which costs billable time the shop never gets back.
- 7:30am
The phone lines open. In a heat wave the CSR is booking, taking payments, fielding where's-my-tech and re-booking yesterday's parts-return jobs at once. After hours there are three options and no fourth: the on-call tech's personal cell, an answering service that takes a message and books nothing, or voicemail.
- 10:40am
The counter. The transaction is: will-call, give them your PO number, confirm, go. The PO number is the only thread connecting a physical part to a job. If the tech says just put it on our account, the thread is cut permanently, and job #4471 will report a $912 invoice, zero material cost, and 100% gross margin. The owner sets next season's pricing from that report.
- Noon
The lunch-hour call gap. Nobody is on the phones, and nobody has ever measured what it costs. The figures being quoted for it are not ours to repeat.
- 1:30pm
A cancellation opens a 90-minute hole, which is pure cost. The documented practice is a standby list of reschedulable maintenance to backfill it. Most shops this size keep that list in one person's head.
- 3:00pm
The tech presents Good/Better/Best. The customer says let me talk to my wife. The estimate becomes an open opportunity; working it is rehash, and the role that does it is a follow-up coordinator. An eleven-truck shop does not employ one. So that revenue simply does not happen.
- 5:30pm
Where the data dies. Timesheets are approximations; on Friday the office manager fixes eleven of them by hand from memory and text messages, then re-types the corrected hours into the payroll provider.
- 7:00pm
The owner does quotes. 9:00pm, the owner does payroll. 11:00pm, the on-call phone rings.
None of that is a software gap in the usual sense. It is an office running on one person's phone and everyone's memory, inside tools that are excellent at what they do and silent about the rest.
The leaks
Four leaks you can measure
This is not a niche problem, and you are not an unusually messy shop. Across plumbing, HVAC, electrical and roofing there are roughly ~220,000 US establishments, and about nine in ten run under 20 people (US Census County Business Patterns 2023 (cbp23us.zip), computed). At that size there is no follow-up coordinator, no procurement desk and no analyst. The leaks below are structural, and each one is measurable from paper you already have.
01
The phone
Across home services, 52% of calls are answered by a person. Sixty-five percent for calls that ring past fifteen seconds, seventy-three percent past thirty. The spread across trades runs from 32% to 74% — which trade you're in moves this number more than how well you run the phone.
Invoca, Lead Conversion Benchmarks Report 2026 — the home-services segment of a ten-industry, 70-million-call dataset.
Two notes, because this number is quoted badly almost everywhere.
It is the home-services number, not the all-industry one. Invoca publishes both. Its cross-industry figure was 61% in 2025 and 56% in 2026; the home-services figure was 55% in 2025 and 52% in 2026. Home services runs five or six points below the average every year. If you find 61% attributed to contractors somewhere, that is the all-industry number wearing the wrong label — the trade press made exactly that mistake with this report.
And “answered by a person” is not “answered.” Invoca counts a call as unanswered here even when an automated system handled the caller perfectly well. This is not a missed-call rate, and we will not multiply it by a dollar figure to tell you what you lost. Anyone who does is selling you arithmetic Invoca's own methodology does not support.
The phone is also a paid-media problem now. Over 90% of Local Services Ads leads arrive by phone, and Google's own ranking documentation says:
“If you regularly fail to answer calls or respond to messages, your ad ranking may be affected.”
Source: Google Local Services Ads help — Google's own ranking documentation · verified 2026-08-01
An unanswered phone costs twice: the call, and the rank.
02
The invoice gap
The best third-party dataset in the trades measures billing capture: hours invoiced against hours logged. The median contractor captures 97.1%. The bottom decile captures 66.9%. Only 46% of jobs are fully invoiced; roughly a fifth sit partial or unclassified, and the bottom decile takes more than 30 days to bill finished work. Level CFO's own reading of its bottom decile: “process breakdown between field work and invoicing.”
Level CFO, 2,200+ contractors, per-metric samples disclosed. Part of the data comes from PE due diligence, which likely skews toward larger, sellable firms.
Source: Level CFO Contractor Benchmarks (updated 26 July 2026) — 2,200+ contractors, all 50 states, per-metric sample sizes disclosed (billing capture n=963) · verified 2026-08-01
03
The unsold estimate
The one good conversion number in the trades is 73.9%, and it is measured on decided quotes only: quotes that got a yes or a no. Estimates that were issued and never answered are excluded. The gap between issued and decided is the leak itself, and it is not in anyone's number. Separately, 47% of contractors above $10M say estimate follow-up generates 11 to 15 percent of income, against 23% of struggling firms doing it at all.
The follow-up figures are ServiceTitan/Thrive Analytics survey data, and the same study claims zero ServiceTitan users are struggling. That is a selection artefact, and it tells you what the sample is.
Source: Level CFO Contractor Benchmarks (quote conversion, n=794) · verified 2026-08-01
04
The supply house
The category leader's own documentation states that it does not track what is on the truck in real time (it does track what needs restocking), and that its three-way match between PO, receipt and bill excludes Non-PO bills. A part grabbed off the van or counter-bought without a PO is exactly the Non-PO case. Downstream, the statement lands with dozens of line items spanning several jobs, and the rational bookkeeper codes the whole thing to Materials with no job attached. The margin report that results is fiction in the direction that always looks good.
Quoted from ServiceTitan's own help documentation, verified 2026-08-01. A statement of product scope, not a criticism. The full quotes are in the platform section below.
Source: ServiceTitan help — truck inventory replenishment (vendor documentation) · verified 2026-08-01
The refused claims
What we won't tell you
The benchmarks circulating in this trade are polluted, mostly by companies selling phone software. These are the numbers we refuse to put on this page.
We won't tell you what percentage of your calls go unanswered. The published range runs from 14% to 62%, for the same year, presented as fact, and every version traces back to a vendor blog.
We won't tell you what a missed call costs you. No credible source exists for that number. We'll measure yours.
We won't tell you your callback rate, your windshield time, or what a truck roll costs you. Nobody knows those for residential trades. The figures being quoted at you come from enterprise field service, or from nowhere.
We won't promise a booking rate. The category advertises 90% and up; the one vendor that publishes both numbers reports 70% overall and 90% “capacity adjusted,” without defining the adjustment. That gap tells you what the other numbers are.
We won't tell you what your marketing should cost as a percentage of revenue. That convention is universal across agency websites, cited by none of them, and every one of them sells marketing.
We won't make earnings claims. In March 2026 the FTC permanently banned the owners of an AI sales-automation company from marketing business opportunities over exactly those claims, on a product sold as replacing human customer service representatives.
We won't build review gating, and we won't text your cold list.
We won't tell you we can automate your warranty claim submissions across every manufacturer. Most of those portals have no API and several block automation. We'll tell you which claims came back short.
We'll run a 20% holdout for 60 days on anything we claim credit for. It makes our numbers look worse, and it's the only way either of us learns the truth.
The number everyone else prints here is wrong, and we can show you why.
The figure in circulation is that 61% of calls to home-services businesses are answered by a person. 61% is real, but it isn't home services — it is Invoca's average across nine industries, and a trade publication attached it to a contractor headline. The home-services number in that same 2025 research is 55%. In the 2026 edition it is 52%.
We use 52%, we label it, and we will not turn it into “48% of your calls are lost.” Invoca counts a call as unanswered even when an automated system solved the problem. The complement of this number is not lost revenue, and a vendor who multiplies it by your ticket price is doing arithmetic the source does not support.
Source: Invoca, Lead Conversion Benchmarks Report 2026 (home-services segment) · verified 2026-08-01
One leading AI vendor in this category currently renders its homepage headline results as unfilled placeholders reading “0% Increase in booking rate.” That is the state of proof in this category. Nobody is auditing these numbers.
The most honest published case study in the category warns that its own figures “aren't measuring the same period, the same season, or even quite the same thing.” We quote it approvingly. The one vendor who undercut their own numbers deserves the credit.
The work
What we actually build
Each row carries its caveat, because a skeptical owner tests us on exactly these, and the caveats are the commercially valuable half. Dashboard fields are named because vague dashboards are how this category hides.
Materials that never reach the job
We pull the supply-house statement, match PO against receipt against bill, auto-post the clean lines and queue only the exceptions: no PO, no job code, price above contract, credit issued but never applied. The tech who bought the part gets a same-day text with the SKU and one question: which job?
On the board: material dollars with no job attached this month; lines with no PO, by technician; credits issued and not applied, aged 30/60/90; jobs reporting margin above your own plausibility ceiling.
The honest caveat: most shops don't issue POs at all. The real first project is instituting them, a behavior change wearing a software costume. And this surfaces theft, not “may”: have a plan before go-live.
Warranty claims, tracked to credit-applied
Data-plate photo required before an install closes; registration task opened against the manufacturer's own window; warranty status pushed to the tech before arrival; claim packet assembled and chased through the whole state machine, down to the credit actually landing on your distributor account.
On the board: installs inside the registration window and not yet registered; approved claim dollars not yet credited, aged; days from submitted to credit-applied, by manufacturer.
The honest caveat: automated submission across every manufacturer is not deliverable today, by us or anyone. Submission stays human. Several brands gate the extended parts term to a registration window (60 days for Trane and Lennox, 90 for Carrier; check your own brands' terms). Deadlines already blown are gone, and the audit says so on page one.
The phone, without a robot deciding anything
Missed-call text-back with a duration floor and a suppression check, sent from the number that was dialled. After-hours intake against your own written emergency list: gas smell, CO, active water and no-heat below your stated temperature always warm-transfer to a person. Every call classified against one written definition set, reconciled through to the invoice.
On the board: answered by a person, by the agent, or not at all, hour by hour on a heat grid; booking rate stated three ways, with completed-to-invoiced as the headline; every emergency classification listed individually, never aggregated.
The honest caveat: a text into an unmonitored Friday-evening inbox is worse than a missed call. A no-heat call at 20°F wants a voice. And we always disclose the automation. It is not a setting.
Estimate follow-up, inside the rules
Open opportunities polled from your FSM, segmented by dollar band, gated record by record against the 18-month relationship window and your consent ledger before anything sends. Short sequence, specific messages, sent as the shop and naming the tech. Any reply halts everything and hands to a person.
On the board: the ghosted set, counted and dollared, which most shops have never seen; objection reasons, ranked, which is the deliverable that improves the proposal rather than the follow-up.
The honest caveat: this amplifies whatever your sales process already is. It cannot fix a price problem or a trust problem, and a shop with a bad proposal gets told no faster. Bereavement and hardship have no data field, which is why every reply stops the sequence.
The invoice gap
A close-out gate per job type: debrief, photos, parts, signature, or the job doesn't close. A nightly unbilled-aging alarm. The collections ladder actually executed, with the day-5 call asking about the service before it asks about money.
On the board: hours logged against hours invoiced, by month and technician; jobs blocked on a missing artifact, by artifact; the ladder stage each invoice is at versus where your policy says it should be.
The honest caveat: the first output is uncomfortable rather than actionable, and it will show that some logged hours are fiction. That fix is management, not software. We won't enforce a gate you haven't agreed to enforce.
The membership base nobody can count
First, a reconciliation project, usually two to four weeks: agreements with no expiry, duplicates, and the one that stops everything, visits sold and not delivered. Then a T-45 summary of what the member actually got, a T-30 card check, and a T-14 call task with the history on screen.
On the board: agreements current, expired, lapsed and unknown (a real bucket); visits contracted against visits delivered; agreements in deficit, which is the do-not-solicit list.
The honest caveat: renewal automation exposes unfulfilled obligations. If you sold two visits and delivered one, the value summary is an invoice for your own failure. We will not automate a renewal ask to a member you owe a visit.
October, planned for in July
Maintenance outreach that reads your board two to six weeks forward and offers only genuinely low-demand slots, members first, by contractual obligation. A live standby list to backfill cancellations.
On the board: your revenue curve and your payroll curve on one chart, from your own P&L; valley fill against peak-booked, where the second number should be near zero.
The honest caveat: we cannot make demand counter-seasonal. We can move discretionary work into the valley. That is a smaller claim, and it is the true one. A campaign that books 200 tune-ups into your first hot week is worse than no campaign.
One page on Monday
Nightly extract from FSM, accounting, telephony, reviews and ad accounts into one warehouse. Six to eight metrics, each against a written definition agreed in a workshop before anything is built. A variance line every month naming where the dashboard and the P&L disagree, and why.
On the board: revenue by business unit, margin, average ticket by department, booking rate to the invoice, callback rate auto-detected, utilization, AR aging. Not forty KPIs. Eight.
The honest caveat: this is the worst first project on this list, and we recommend against buying it first, in writing. Reporting on broken processes produces accurate pictures of chaos. And we will not sell real-time job profitability; materials land days later on a statement, so we ship trailing materials and call it that.
Permits that stall the job
Permit-requiring work flagged at sale by trade and jurisdiction; install scheduling blocked until issued, overridable only by you and logged when it is; a status chase with alarms; and the register of open permits by year, the one that turns up at a property closing.
On the board: sold, permit required, not yet applied for, aged; applications pending beyond your threshold, by jurisdiction; open permits never closed out, by year.
The honest caveat: there are roughly 20,000 permitting authorities in the US and most have no API. We will not build universal portal automation, and we will not quote a coverage percentage we cannot measure in your metro. We sell visibility, and we say the word. The system may warn; it never authorises.
The refused products
What we decline to sell first
The refused-claims list above has a sibling: things we're asked for and won't build first, or at all. No competitor publishes one, which is rather the point.
Technician scorecards and leaderboards
The automation most likely to blow up a company. Rank on average ticket in a tight labor market and you get overselling and eventually a board complaint; mix bias makes good techs look bad, and they quit. Every scorecard needs a counterweight metric, and none of that is a month-one build.
Commission and spiff automation
The highest-emotion pain in the business: the one where a mistake means a tech quits. We surface the inputs. We do not compute the cheque.
Dispatch optimization
Wrong answer below roughly six trucks, and above that the dispatcher is often right in ways no algorithm encodes. Any system that enforces the paper priority ladder will be worked around within a week. What we will build is the scoring inputs: close rate and actual duration by technician and job type, from your own history.
Texting your 8,000 names
The highest-legal-risk item in the category. Outside the 18-month relationship window it runs $500 to $1,500 per message in theoretical exposure, against a list containing reassigned numbers, and the STOP-rate spike degrades the appointment confirmations your dispatch runs on. Outside the window, we refuse in writing.
Recruiting automation
Automation doesn't create candidates, and faster screening of zero applicants is zero. Knockouts stay on objective credentials: license, EPA card, driver's license, documented. For a shop hiring one or two people a year, this is the item we talk you out of.
Review gating, in any form
Google prohibits it, the FTC Reviews Rule makes suppression a civil-penalty matter, and the Google API cannot request reviews at all. We ask every paying customer once, unconditionally, on paid invoice. On Yelp, we don't ask.
Worksheet
The arithmetic, on your numbers
Every field starts empty, because defaults are claims. Every number that touches the arithmetic is on this page and editable, because hidden coefficients are claims too. And it can return an unimpressive answer: below roughly $1.5M in revenue or 150 calls a month, our judgment is that most of this does not pay back and the better spend is a good CSR. That threshold is our judgment, not a benchmark.
Nothing you type leaves this page. There is no request, no storage and no analytics on these fields; you can verify that in your browser's network tab.
Schedule A · the phone
Pull last month's call detail record from your carrier. Benchmark context sits in the leaks section above, cited; none of it is pre-filled here.
- 1
- 2
- 3
- 4Ceiling on what those calls could have been worth(Line 1 × Line 2 × Line 3)
Line 1 × Line 2 × Line 3 = —
A ceiling, not an expectation. Some of those callers called back; some were never going to book; nothing recovers all of line 1. Any vendor modelling this as recovered revenue is telling you a story.
Schedule B · materials with no job attached
Pull last month's supply-house statements and the same month's invoiced revenue.
- 5
- 6
- 7Share of the month's material cost sitting in overhead instead of on a job(Line 5 ÷ Line 6)
Line 5 ÷ Line 6 = —
Not a savings figure, and we will not convert it into one. It measures how wrong your margin report is. Context, someone else's data: in the largest contractor dataset available, 91% of jobs carry revenue with no cost data attached (Level CFO, 2,200+ contractors).
For scale, beside your result rather than instead of it: Corveonic retainers are published at $597, $1,197 and $1,997 a month plus a one-time setup fee.
Client-side only. Your numbers never leave your browser; we never see them.
Benchmark context above: Level CFO Contractor Benchmarks — 315,393 jobs across 1,391 companies · verified 2026-08-01
The incumbent stack
You already pay for most of a platform
ServiceTitan, Housecall Pro and Jobber are excellent inside themselves. ServiceTitan ships AI answering, customer self-booking, AI dispatching and an agentic layer; Jobber sells an AI Receptionist for $29 a month. We will not tell you these platforms can't do things they can do. We work the seams they leave, and every seam below is quoted from the vendor's own documentation so you can check it.
The $29 figure: Jobber published pricing (AI Receptionist add-on; free on the Plus plan) · verified 2026-08-05
The sharpest fact available here comes from a competitor's SEC filing. ServiceTitan prices per field technician; the 10-K's own words:
“generally based on the number of field technicians”
Source: ServiceTitan FY2026 Form 10-K, filed 25 March 2026 · verified 2026-08-01
Their bill grows every time you hire. Automating office work doesn't. The same filing lists TitanAdvisor, a shipped product whose stated job is to recommend features customers already pay for and haven't turned on. A company that builds that product is telling you what the binding constraint in this market is: adoption and configuration, not features.
The truck
ServiceTitan's help documentation, in full, because the second sentence matters:
“ServiceTitan does not allow you to track on-hand inventory in real-time (total inventory you currently have on the truck). However, you can track which inventory needs to be restocked in real-time.”
Source: ServiceTitan help — truck inventory replenishment (vendor documentation) · verified 2026-08-01
Restock-need is tracked in real time; on-hand quantity is not. You cannot price a job against inventory you cannot count.
The bill
Three-way matching between PO, receipt and bill is gated behind a procurement tier reached through a sales address, and the documentation states that it
“does not include Non-PO bills”
Source: ServiceTitan help — 3-way matching (vendor documentation) · verified 2026-08-01
A part counter-bought without a PO is exactly the Non-PO case. The highest-leakage path is the one the tool explicitly excludes.
The ledger
The FSM matches QuickBooks customers on exact name at first export,
“including commas and spaces”
Source: ServiceTitan help — how ServiceTitan syncs with customer records in QuickBooks Desktop · verified 2026-08-01
Any mismatch silently creates a duplicate customer, with no documented remediation. The structural reading is the honest one: the FSM and the ledger share no key. They reconcile on strings, which is why this class of failure never fully goes away.
The permits
For where permit research should live, the category leader's own guidance is
“a three-ring binder, a Google drive, a ServiceTitan content portal or a shared drive on your company's network”
Source: ServiceTitan — pulling permits (vendor guidance) · verified 2026-08-01
A binder, recommended inside a platform priced per technician. For tracking pulled permits, the same page separately suggests a shared spreadsheet. Both recommendations are real, and they are separate.
And the lead market, on the record
Angi's own FY2025 filing defines that a single service request can produce multiple leads: about 1.3 per request on the filing's own figures, a derived ratio and the honest one (the complaint that it is four is not supported). The FTC, separately, ordered HomeAdvisor to pay $7.2 million over lead-quality misrepresentation, an order HomeAdvisor notes does not admit or find wrongdoing. None of that means stop buying leads. It means the phone you answer is the asset, and the leads you dispute are worth disputing on time.
Source: Angi Inc. FY2025 Form 10-K, filed 20 Feb 2026 (derived: 20.2M leads ÷ 15.5M service requests) · verified 2026-08-01
Source: FTC order against HomeAdvisor, Inc. (Docket 9407; final order 21 April 2023, vote 4-0) · verified 2026-08-01
Competitor prices and terms on this page were verified 2026-08-01 and are treated as expiring: re-verified before launch, then quarterly. The access date travels with the figure.
Compliance
The rules we build to
Not downplayed, and not a scare section. These are constraints we build in before you ask, because the alternative is a vendor who finds out about them from your lawyer.
Texting has federal rules, and the product enforces them
Quiet hours are 8am to 9pm at the customer's local time, built as a hard, non-overridable send window, because “text two hours after job complete” fires at 9:45pm on a 7:45pm job. Opt-outs are honored across every automation from one shared suppression ledger, and the records are kept. Exposure for getting this wrong runs $500 per message, trebled to $1,500. One rule you may have read about is no longer live: the one-to-one consent rule was vacated in January 2025, while the 2012 written-consent rule stands. Many vendor pages still describe the vacated rule as current.
We will refuse to blast your old list
The established business relationship runs 18 months from a transaction. A four-year-old unsold-estimate list is outside it by definition, and the STOP-rate spike from a cold blast degrades the carrier trust score your appointment confirmations depend on. The list tells you who to call. It does not tell you that you may.
Review gating is prohibited, and we won't build it
Google's policy bars selectively soliciting positive reviews; the FTC Reviews Rule makes suppression a civil-penalty matter; and Yelp's rule is the opposite of Google's: don't ask at all. We ask every paying customer once, unconditionally, on paid invoice. You will also get more one-star reviews. That is the deal, and any vendor who doesn't say so is setting you up.
Recording and disclosure
New York is one-party consent; the customer's state may not be. We assume all-party consent, disclose at the head of the call, and disclose any automation that speaks or writes to a customer. Always on. Not a setting.
The licensed human decides
A diagnostic suggestion is an assist, not an authority. The person who opens a sealed system holds a federal EPA Section 608 certification; the software does not, and never decides. In-home sales carry a federal three-day right to cancel, and the emergency-waiver branch is exactly the decision an automation never makes alone.
The paperwork is the defense
The TCPA provides an affirmative defense for reasonable practices and procedures, implemented with due care. The consent ledger, the suppression list and the audit log are not admin. They are that defense, and they ship at every tier.
The texting figures: 47 U.S.C. § 227; 47 CFR 64.1200 (TCPA and FCC rules) · verified 2026-08-01
The vacated rule: Eleventh Circuit, Insurance Marketing Coalition Ltd. v. FCC, No. 24-10277 — “PETITION GRANTED; ORDER VACATED IN PART AND REMANDED” · verified 2026-08-01
The review rules: Google review policy and Yelp policy, verbatim (PL §9.1); FTC Reviews Rule, 16 CFR Part 465, effective 21 October 2024 · verified 2026-08-01
The certification: EPA — Section 608 technician certification (federal regulation) · verified 2026-08-01
What happens to client data is stated in plain language in our privacy policy.
Read the privacy policyThe season
July pays for October
Watsco is North America's largest HVAC distributor and a public company, so its quarters are on the record: Q2 and Q3 carry 57.0% of full-year revenue, and the peak quarter runs 35% above the trough.
A distributor's mix includes commercial and new construction, which damps the swing; a residential service contractor's curve is sharper. Treat 35% as a conservative floor, never as your own number.
“Demand related to the residential central air conditioning replacement market is typically highest in the second and third quarters, and demand for heating equipment is usually highest in the first and fourth quarters.”
Source: Watsco, Inc. FY2025 Form 10-K, Item 1 “Seasonality” (filed 27 Feb 2026) · verified 2026-08-01
Now the half that matters more: headcount barely moves. Federal employment indices put the peak-to-trough employment swing for plumbing and HVAC shops at 4.6%. Revenue swings by more than a third; payroll moves by less than a twentieth. That gap, payroll flat while revenue collapses, is the October problem, and it is the actual argument for an office layer that flexes without hiring or layoffs. The administrative work that smooths the season is crowded out by the season.
Source: BLS CES employment indices (series CEU2023822001 et al.), computed from the BLS public API, 2019–2025 averaged, data through May 2026 · verified 2026-07-31
Source: Watsco, Inc. SEC filings: Q1/Q2/Q3 FY2025 Forms 10-Q (accessions 0001193125-25-116461, -176973, -272272) and Q4 via Form 8-K Ex. 99.1 (0001193125-26-053048) · verified 2026-08-01
FAQ
Asked by owners, answered without a script
Will this replace my CSR?
No, and if a vendor says yes, ask what happens on the third 100°F day in July. Industry-wide, only 52% of inbound home-services calls reach a person, and 55% of businesses never ask the caller to book. Your CSR isn't the problem; she's outnumbered. And on the market generally: 75% of consumers say they prefer a human for customer service, and Klarna publicly reversed an AI-only support strategy and re-hired people. The design that works is augmentation with clean human escalation.
I already pay for ServiceTitan / Housecall Pro / Jobber. Isn't this duplicated?
Often, partly, and we'll tell you where. If turning on what you already own solves it, do that first; we'll scope which ones free. We work the seams: supply-house PO reconciliation, warranty claims tracked to credit-applied, permits, and reporting that reconciles to your QuickBooks P&L using your burden rate. Your platform's own documentation says it does not track on-hand truck inventory in real time (it does track restock needs), gates three-way match behind a procurement tier that excludes non-PO bills, and recommends a binder or a shared drive for permits. Those aren't insults. They're published.
Is texting my customers legal?
Yes, with real rules, and one you may have read about is no longer among them: the one-to-one consent rule was vacated in January 2025, while the 2012 written-consent rule stands. We build the quiet-hours window, the opt-out handling and the record-keeping into the product from day one, and the data-handling picture is stated in our privacy policy.
Read the privacy policyCan you text my whole list to drum up work?
No, and we'd push back if you asked. The established business relationship runs 18 months. Beyond that you're looking at $500 to $1,500 per message in theoretical exposure, and, more immediately, a STOP-rate spike that lowers your carrier trust score and starts degrading the appointment confirmations your dispatch depends on.
Can you get me more Google reviews?
We can ask every paying customer once, unconditionally, with no incentive. We won't gate: Google prohibits it and the FTC Reviews Rule makes it a civil-penalty matter. And you'll get more one-star reviews too. On Yelp, the correct answer is not to ask at all.
How long until something works?
First automation live in two to four weeks, and it'll need no new behavior from your techs. Your data sets the timeline, not our code: cleanup is typically most of the project, not a preliminary to it. We'll tell you which you are in week one.
How will I know it worked?
Baseline from your system, in writing, before we build. One metric per automation, measured at the invoice, not the lead. And a 20% holdout for 60 days, which makes our numbers look worse and is the only way either of us learns the truth.
What if we part ways?
Everything runs in accounts registered to you, and your phone numbers and 10DLC brand registration stay in your name. That's where firms get trapped: if the agency owns the number, leaving means losing the number painted on your trucks and rebuilding carrier trust from zero. Ask any vendor who owns the registrar, the numbers and the brand registration, and get the answer in writing.
What this costs: retainers run $597 to $1,997 a month plus a one-time setup fee, published in full, with no call required.
See the tiersOr run your own numbers firstWe'd rather measure than promise.
Thirty minutes. Bring last Thursday's call log and one supply-house statement, and we'll tell you which leak to close first, or that you're too small for this to pay back yet. Both answers are free.