ROADTRUTH monitors every vehicle in a fleet through the telematics it already carries, computes a Vehicle Health Score (VHS, 0–100) for each unit, predicts what is about to break inside a time window, and then holds itself to account: every prediction is adjudicated against what the workshop actually found, and both confirmations and misses are published on the Prover Board.
The platform is organized like a clinic. AI Physicians read each unit's case file (fault codes, telemetry, service history) and draft prognoses; a human maintenance lead reviews and posts them; the shop's findings close the loop. Three honesty rules run through everything: simulated data is always labeled, cost figures are estimates and never quotes, and nothing posts to the board without a human confirming it.
ROADTRUTH is paired with ROADWORTHY, its driver-readiness sibling. Vehicle data flows to the driver's app; only a 0–100 readiness aggregate flows back. Driver health never crosses the wall.
The Floor is the fleet at a glance — one card per unit.
Rounds is the floor's conversation: the Physicians walking the fleet, challenging each other's readings, answering the Adversary. Use ASK THE FLOOR ⚡ to put a question to the live AI channel — "what's the risk on LAG-07 this week?" — and the floor answers from the actual case files.
The board is the platform's public record. Every prediction appears with its type, interval, and status:
| Status | Meaning |
|---|---|
| PENDING | Posted; the interval clock is running. |
| CONFIRMED | The shop's findings matched the prediction. |
| MISS | The findings refuted it. Misses stay on the board — they train the engine. |
| WITHDRAWN | Pulled before adjudication. |
Statuses flip when a teardown report arrives (see WhatsApp shop reports).
The ninth agent, powered by Claude. RUN AI FORECAST on any unit produces a 90-day outlook: per-item risk, expected window, recommended action, and a cost range grounded in your fleet's own logged invoices — plus a quarterly maintenance budget figure. Every figure is marked ESTIMATE · NOT A QUOTE, and each item states its basis. The more service records the Office logs, the sharper the cost grounding becomes.
The Office is the fleet maintenance desk, gated by PIN. The master Office PIN sees everything; a fleet PIN (set per carrier) scopes every action below to that carrier's own units — pairing, service records, predictions, import, registry view, and reports.
Enter the driver's DoktorConnect ID (DC-XXXX-XXXX) against a unit and PAIR. The driver's "My Truck" card and Lane readiness strip light up in ROADWORTHY on their next open.
LOG WORK after every job — description, parts, odometer, hours, cost, currency. This is the Forecaster's cost reference base: the fleet's own invoices become the pricing truth.
No hardware is installed on any vehicle at any step. Access is read-only, scoped, and revocable by the carrier at any time.
After a repair, the mechanic sends one short WhatsApp message to the fleet's ROADTRUTH number: the prediction id (or the unit), what was found, and whether it matched — a photo helps. The AI parses the message, matches it to the open prediction, and flips the board to CONFIRMED or MISS. Example:
P-0147 LAG-07 — opened the alternator, rectifier diode pair burnt exactly as predicted. Replaced. Confirmed.
Report the misses with the same energy as the wins. A miss on the board is the engine learning; a miss kept quiet is the engine lying.
| You see | It means |
|---|---|
| "Awaiting first telemetry window" | The unit was just imported; the next 15-minute ingestion pass will populate it. |
| ENV FALLBACK on a fleet | The fleet has no stored credential of its own; it rides the deployment's environment token if one is set. |
| A score changed after a repair | Correct behavior — the engine recomputes from current evidence. Adjudicating a prediction removes its risk penalty on the next pass. |
| DEMO label on output | An honest fallback: the underlying service or table isn't configured yet, and the platform says so instead of inventing data. |
| "This pin is scoped to another fleet" | You used a fleet PIN on a unit that belongs to a different carrier. Use that fleet's PIN or the master PIN. |
| AI PROPOSE returns nothing | The case file was too thin to justify a prediction. Wait for more telemetry or fault evidence. |
Josmol Group Inc. · Joseph@josmolgroup.com · 915-478-8807 · 4816 N. Mesa St, El Paso, TX 79912