Franchise systems · 3–30 location home service operators
Ask AI about
your locations.
It gets some
of them wrong.
Your brand has twenty front doors. Twenty Google profiles, twenty sets of hours, twenty phone numbers — most of them claimed years ago by a franchisee who has since changed one of them and told nobody. When a customer asks an AI engine about your brand in a given city, the answer is assembled from whichever of those is worst.
Nobody at corporate is looking at this, because there is no report that shows it. We go look — location by location, logged out, with dated screenshots — and hand you what it actually says.
Asked of a major engine, logged out:
"Who should I call for emergency AC repair in [city]?"
Three of those four are quiet. Nobody complains about a location an engine never mentioned.
Why it happens
Nobody owns the gap
between the brand
and the listing.
This isn't neglect. It's structural, and it's specific to how franchise and multi-location systems are actually built. Four parties, each doing their job correctly, and the result still drifts.
Owns the brand and the website
Sets standards, supplies the templated site and the locator. Has no login to the franchisee's Google profile and no report that would show a wrong number.
Owns the listing
Claimed it themselves, often years ago, sometimes on a personal account. Changes hours or a number when the business changes and has no reason to think of it as a brand asset.
Keep the old version
Data propagates outward and stays there. A number corrected in one place lives on in the eleven aggregators that copied it.
Assemble from all of it
They read every version and answer with confidence. Where the sources disagree, the brand tends to get skipped in favour of a competitor whose sources agree with each other.
The platforms that solve this at scale — the ones with dashboards and annual contracts — are built and priced for brands with hundreds to thousands of locations and a marketing department to run them. At three to thirty locations you're beneath their sales floor and above what a local SEO freelancer can do across a system. That's the gap this firm was built in.
The other half
And then there's
the phone.
A wrong listing costs you the customer who never called. The phone costs you the one who did. Both are uncounted, both show up in the same audit, and this is the half that already has a formula.
12 locations × 4 missed/day × 30% close × $600 ticket × 250 operating days
Every assumption in that default is deliberately conservative, and you should replace all five with your own. The structure is the point, not the number.
Worth saying plainly: we are not primarily a phone company, and if your phones are handled we will tell you the call census came back clean rather than sell you a system you don't need. Plenty of franchise systems already have an approved answering vendor. Almost none of them have anybody checking what an engine says about location eleven.
Missed calls don't appear on any report you currently receive. There's no line item for them. No location manager is going to raise a hand in a Monday meeting and say they let four a day go to voicemail in August. So the figure stays invisible right up until somebody pulls the call logs location by location — and then it's usually the largest uncosted line in the business.
Run your own numbers before anybody sells you anything, including me.
Step zero · the AI Operations Audit
Most of the first hour
is counting things
nobody counts.
Two halves, one fixed scope: what people find when they look for each of your locations, and what happens when they call it. At three or more locations neither one is ever what the dashboard says, because for most of it there is no dashboard. The audit produces a written document whether or not you implement anything with us.
At Innovo, the first hour of that kind of review surfaced $11,000+ in uncollected subscription billing and an $18,000/month dormant delivery line. Neither was hiding. Both were simply uncounted.
The front door
Before the phone, what a caller finds. What the AI engines say about each location when somebody asks for your trade in that city — logged out, dated screenshots, run per site rather than once for the brand. Then the mechanical layer underneath it: hours, addresses and numbers that disagree between your own site, your listings and the map; click-to-call links that don't dial on a real phone; forms that fail silently; location pages that don't exist. For public agencies this pass also runs WCAG 2.1 AA against the Title II deadline, and the findings come back in that language.
Answering the phone faster does nothing if the phone was never going to ring. This is why it's step one.
The call census
Every location's call log, pulled and counted the same way — answered, missed, abandoned under fifteen seconds, after-hours, repeat callers. One method across all sites, so the numbers can finally be compared to each other.
The intent inventory
What people actually call about, by location and by hour — service, quote, billing, delivery, emergency, hiring, and the ones nobody has a bucket for. This is what routing gets built from, and it's the step most vendors skip.
The routing map and the leak review
Where a call goes today versus where it should go, drawn out — including what happens after hours and what happens when the first person doesn't pick up. Alongside it, the revenue review: subscriptions, recurring work, and standing lines that stopped being billed.
The findings document
Written findings sorted Now / Next / Later, split into what we would do and what your team does, with a scoped implementation and a real number attached. Every finding traces to something countable — no finding rests on an impression.
$10,000, fixed. The document is yours either way — implement it with us, hand it to your own operations team, or put it in a drawer. Plenty of people stop here, and that is a legitimate place to stop.
Franchise-ready
One corporate line.
An instance per location.
"The AI handles it" is not an architecture. Make anyone selling you this draw it. Here's ours — the same structure Innovo's own franchise build uses, which is why it exists at all.
Innovo's build is structured exactly this way: a corporate number at the top, and each location holding its own sub-account, its own local number, its own instance, and its own territory zones. Adding the twentieth location is a configuration, not a rebuild — which is the entire reason a franchisor can price it.
The Glass Box standard
The failure mode is
never a dramatic outage.
It's a small wrong answer, delivered confidently, forty times before anyone notices. At one location somebody eventually mentions it. At twelve, nobody does. These four controls are non-negotiable on every install for exactly that reason.
Transcript after every call
To ownership, every call, not a sample. If nobody is reading them, they're decoration — so the audit also names who reads them and when.
Mandatory customer lookup
The record gets pulled before anything is said. A ten-year account is never treated as a stranger, and nobody is quoted a price that belongs to another location's job.
A human escape hatch
A defined handoff with the trigger conditions written down rather than assumed. Knowing when to stop is a specification, not a shortcoming.
Emergencies to on-call cells
Anything urgent leaves the system and reaches a phone in somebody's pocket. That path is tested on a schedule, not hoped for.
The boring parts
SLAs, incident response,
and who gets called at 2 a.m.
This is the section most vendors don't write, and it's the one that decides whether a system survives contact with twelve locations, a franchise agreement, or a public agency's procurement office.
Incident response, written down
What happens when the system can't reach on-call. Who is called second, how long it waits, and what the caller hears in the meantime — decided in advance, in writing, instead of improvised at 2 a.m.
Change control
Every fix becomes a written rule the system keeps, and rules are versioned. That's how a correction at one location stops being a correction anyone has to remember to repeat at the other eleven.
Redaction standard
Customer names, numbers, and identifying details are redacted before a transcript is used for anything outside operations — review, training, or a case study. Client recordings are never used publicly without written permission.
Staff training at cutover
Location managers and dispatchers are trained on what the system does, what it hands off, and how to override it — before it answers a live call, not after the first complaint.
On public-sector work, plainly: we're early. The operating controls above are in place and documented, and we will describe them to any procurement officer who asks. What we won't do is print a capability statement listing contract vehicles and past performance we don't yet hold. Ask us what we have; you'll get the real answer.
If you're a city, county, district or agency
Your website has a federal deadline, and it has a date on it.
Title II requires WCAG 2.1 Level AA of state and local government web content and mobile apps. April 26, 2027 if you serve 50,000 or more; April 26, 2028 if you serve fewer, and for every special district. The front-door half of our audit is that conformance review, written against the success criteria so your vendor can act on it.
Where this came from
I own a service company.
It runs here first.
My husband and I own Innovo Spa — hot tub and swim spa concierge service across Michigan and South Florida. Our own phones run on this architecture, our own forensic review is where the method came from, and our franchise build is why the per-location structure exists.
The straight answer on those two dollar figures: they're operations-audit findings, not results the phone system produced. Counting found that money, not automation. I include them because counting is the job — most of what's broken in a multi-location service business is something nobody has counted lately, and the phone is the biggest one.
What it costs
Published, so you can decide
before the call.
Everyone starts at the audit — including the people who never go past it, and the ones who turn out not to need us.
The audit is 0.5% of the number you just produced. If it's wrong by an order of magnitude in your favour, it's still 5%.
That's the whole argument, and it's arithmetic rather than persuasion. The reason to run the numbers first is so the price stops being a question of whether it feels like a lot of money and starts being a question of whether the figure above is real. Find out for $10,000 or keep estimating for free — those are the two options, and plenty of operators reasonably pick the second.
AI Operations Audit
Where everyone startsThe census, the intent inventory, the routing map, the leak review, and a written findings document with a scoped implementation attached. What's in it.
Implementation
Scoped by the audit, not by a brochureRouting, per-location instances, escape-hatch rules, incident response, booking integration, and staff training through cutover. One time.
Managed operations
Ongoing, after cutoverWe read the transcripts, tighten the rules as calls come in, run the incident tests, and report on what changed. The alternative is that nobody does, and the system quietly drifts.
What lands on your desk at the end.
Five documents, dated, with the working shown. Not a presentation.
The front door
What every engine and every listing says about each location, with dated screenshots, plus the mechanical fix list — dead links, wrong hours, broken forms. In WCAG 2.1 AA terms where you're a public body.
The count
Every location's call volume, counted the same way, side by side for the first time — with the raw pulls attached so your team can check the arithmetic.
The routing map
Where a call goes today against where it should go, drawn out, including after-hours and no-answer paths.
The findings
Now / Next / Later, split into what we would do and what your team does. Every finding traced to something countable.
The scope
A fixed implementation number — not the range on this page — with what's in it and what isn't.
Why the number is what it is.
Straight answers
The five questions
I actually get asked.
What do we actually get for $10,000?
Five written documents, not a slide deck. The front-door review — what every AI engine and every listing says about each location, with dated screenshots, and the mechanical faults underneath it. The call census, every location counted the same way. The intent inventory. The current routing map against the one we'd recommend, plus the revenue-leak review. And a Now / Next / Later fix list split into what we do and what your team does, with a scoped implementation and a real number on it. You own all five whether or not you go further.
Does this replace our people?
No, and any vendor who says yes hasn't run a service business. It answers the calls nobody was answering — after hours, the third line during peak, the ones that go to voicemail while your dispatcher is on the phone. There's a defined handoff to a person, with the trigger conditions written down. The people you have stop losing their day to calls that should never have reached them.
What happens when it gets something wrong?
You see it, because every transcript comes to ownership after every call. You say what you'd have wanted said, and it becomes a versioned rule the system keeps — at every location, not just the one that caught it. That loop is why failures here get measured in days instead of quarters. A real one from our own build: transfer targets were set to desk numbers, which at 2 a.m. is a voicemail box with a professional greeting. It surfaced in transcript review rather than from a complaint. The targets moved to direct cells, a mandatory pre-transfer lookup and a repeat-caller escalation rule went in behind it, and testing that path became a scheduled item in the incident-response protocol.
We're two locations. Is this you?
Probably not, and I'd rather say so now. This firm is built for three or more locations, franchise systems, and public agencies — the controls, the documentation, and the price all assume that. One or two locations with up to about ten trucks is a different job with a different price, and it goes to Lungo Marketing. Same owner, same standard, different room.
How long does an install take?
Send the location list and you'll get a real date rather than a comfortable one. The front-door half is fixed-scope and scales predictably with the number of locations. The build is the part that varies, and it varies with how many booking systems are in play and how much of the routing already exists in writing.
How this starts
Send us your
location list.
Not a discovery call, not a capabilities deck. Send the list of your locations and we'll run the first pass on three of them at no charge — logged out, dated, exactly the way the full review runs. If all three come back clean, I'll tell you that, and it's a cheaper answer than the alternative.