Channel 03 · How it works
One interview in. A station out.
A LocationCast location is built once — presenter, channels, facts — and then run weekly by a ten-stage system in which humans touch only the exceptions. This page is the whole machine, with nothing left in a black box.
Anatomy · Part 1
The presenter: one face per location
Each location is fronted by an AI presenter cast to its trade’s working-owner archetype: the person a customer hopes pulls into the driveway. Casting follows a written standard — face age and voice age must agree, wardrobe and setting match the trade and the season, and the read is a working owner’s, not an announcer’s.
The same presenter fronts the flagship episode, the shorts, and the audio editions, so a town learns one face the way it learns a jingle — except this one can say this week’s offer, this town’s name, and this season’s warning.
Two rules never bend. The presenter is disclosed as AI on the content itself — no synthetic person is ever passed off as a real employee. And no real person’s likeness or voice is used without a documented release; customer stories are told only from real, consented reviews.
Anatomy · Part 2
The channels: a station, not a folder
Every location gets its own YouTube channel, its own social profiles, and its own 24/7 TV channel looping its programming — the same pattern the studio already runs around the clock for its Erie Spotlight channel. Programming is scheduled like a station’s: flagship episodes anchor the week, shorts rotate through the day, seasonal pieces replace stale ones on a calendar, not on someone’s memory.
Accounts are created in the location’s name and held by you (channel 05). The system operates them; it does not own them.
Where that stands today: that is the built-once design. In the pilot it exists today as one studio pattern channel; standing it up per location, at cadence, is what the season tests.
Brand control across a group
A group gets one brand system — presenter standards, graphic language, offer rules, claim rules — applied by machine at every location, which is precisely where human franchises leak. What varies is the facts: prices, towns, offers, team. What never varies is the standard. A location cannot go off-brand, because a location cannot reach the render button; it can only update its facts, quarterly, through the interview.
Anatomy · Part 2, continued
The map pack: each location’s Google profile
For home services, “near me” is decided in Google’s map pack before any video is watched — so a location’s Google Business Profile is part of the station, and it is run like one. The profile is built from the same facts file as everything else: the DBA-correct name, the categories, the service towns, the hours. The flywheel feeds it on the same weekly cadence — finished pieces become profile photos, the week’s offers become Google posts carrying their true expiry dates — and its status (verified, current, dark) is a line in the monthly report, like any other channel’s.
Reviews stay the customers’. The system never writes, buys, filters, or gates a review. A real, consented review can be carried as content under the same rule that governs reviews everywhere in this system: presented as a review, never strengthened.
Where the pilot stands, plainly: setup packets are built for all nine pilot brands — names, categories, phone numbers, hours, and the 13-town service area each one serves. Verification is a step only the owner can complete inside Google, and as of this build no pilot profile is verified live. Until one is, the map pack is a packet on a desk, not a surface we claim — and the pilot’s own scorecard counts it exactly that way.
Anatomy · Part 3
The facts file: the only source copy may draw from
Built from one interview
At launch, each location sits one structured 60-minute interview, drawn from the studio’s question banks — 53 questions per niche, banks built for ten niches. The answers become a facts file: thirty core fields (rates, service towns, hours, guarantees, team, seasonal offers) plus trade modules for coupons, menus of services, codes and permits, seasonal pricing.
The interview repeats quarterly, in shorter form, to refresh offers and seasons. That is the entire recurring time commitment of owning a station: one hour, four times a year, per location.
Claims discipline, in writing
Every claim in every script must trace to a line in the facts file. Prices are typed — a price is fixed, market, or absent, and copy treats each correctly. Every offer carries its own expiry; a flyer-wide date is never smeared across offers that don’t have one. Any savings claim must survive arithmetic against the location’s own rate card before it is scripted.
We institutionalized that rule the hard way: the studio’s own first pilot episode carried a savings line its rate card could not support — the card’s real ceiling on that claim is $575, and the episode said four figures. The episode is held from publication until it is corrected, and the arithmetic check that caught it now sits between every facts file and every script, for every location, permanently.
The copy’s spine
Structure comes from a second asset: dossiers distilled from a 1,709-file archive of proven direct-mail controls — the packages that ran for years because they worked. Offer construction, guarantee frames, itemized value stacks, involvement devices: the dossiers supply the shape, the facts file supplies the substance, and nothing from the archive is ever republished. The distillation is underway now, best material first, and it deepens every season as the system learns which structures produce calls in which trades.
The weekly flywheel
Ten stages. Humans touch stage 10.
Each location, each week, the same rotation. Every stage writes evidence of its own work; a stage that cannot prove its effect shows red on the board instead of being assumed healthy.
| Stg | Stage | What happens | Human touch |
|---|---|---|---|
| 01 | Facts check | The week’s topics are drawn from the facts file and the seasonal module — offers in force, expiries respected, nothing stale. | None |
| 02 | Scripting | Flagship script and short scripts written on dossier structures, claims traced to facts, arithmetic check applied. | None |
| 03 | Production | Presenter performance, plates, captions from the real transcript, graphics — assembled by the production pipeline. | None |
| 04 | Independent QC | A gatekeeper who did not make the piece grades it — claims, captions, casting, arithmetic. It holds bounce authority: fail means it does not ship. | Exceptions only |
| 05 | Approval clock | The batch posts to your portal. Publishing-tier items carry a 72-hour clock; other content, 5 business days. Comment and only that item reopens; silence ships. | A glance, if you want |
| 06 | Distribution | Pieces publish to the location’s channels and join the TV rotation; offers rotate out on their true expiry dates. | None |
| 07 | Timelines | Every deliverable is filed with its editable timeline (OTIO/FCPXML) and masters in your vault — re-editable by anyone, forever. | None |
| 08 | Broadcast week | The station runs: episodes, shorts, and the 24/7 channel carry the week’s programming, every channel carrying the location’s tracked number. | None |
| 09 | Measurement | Tracked calls accrue per location and per channel into the location’s ledger, counted per brand, per week, from the call records — the raw material of the monthly report. Credit stops at brand-week: no mechanism today ties an individual call to an individual piece, and the report never pretends otherwise. | None |
| 10 | Exceptions | Failed renders, dark channels, stale facts, a location with zero calls: each raises a loud exception to a human. This is the only stage with your name on it — and it is ours, not yours. | Our operators |
The week’s yield per location: 1 flagship episode (3–5 minutes), a set of shorts, audio editions, and the graphics set — every week the season runs, including the weeks you are on a roof.