Install Field Report · Roadrunner Septic.
PH—006
Trade services is not fintech. The install proved that the ontology filter works as hard on a septic booking flow as it does on a compliance workflow. Different surface. Same spine. Here is what held, what broke, and what the install taught the corpus.
I What We Installed
Roadrunner Septic operates in a regional home-services market. The install scope was narrow: a booking agent, a scheduling optimizer, and a light intelligence dashboard for the owner-operator.
Install ran seven days. Three capabilities came online. The fourth, Cost Analysis, was skipped by design. For a business this size, three costed scenarios was overkill. One scenario modeled against baseline was enough.
II What Held
The ontology filter held. We had worried that the home-services filter, at that point only three installs old, would miss something obvious. It did not. A recommendation to up-sell customers on a septic service during the booking flow surfaced in the ranked stack. The filter flagged it as a trust violation for trade services specifically. The recommendation was suppressed.
The readiness score held at seventy-six and moved two points up over the first week as new booking data came online. Exactly the movement the ontology predicted.
The agent builder held. Three agents compiled. One for intake, one for scheduling, one for post-service follow-up. Total agent orchestration: eleven steps. All simulated. All signed off.
III What Broke
The dashboard broke on day three. The owner-operator opened it twice and stopped. The screen density was wrong for the user. We had calibrated for a leadership-team reader. The actual reader was one person running a truck.
We rebuilt the dashboard inside the install window. Single screen. Four numbers: calls today, bookings today, revenue today, anomalies flagged. It has been opened every morning since.
IV What It Taught the Corpus
Two signals went back to the ontology.
First, the home-services filter now has a trust-violation rule against same-session upselling on the booking flow. Three installs later, this has suppressed six recommendations that would have shipped in an earlier version.
Second, the dashboard density heuristic now branches by org size. Leadership-team reader versus operator-reader is now a first-class distinction in the ontology. It has changed the dashboard on every install since.
Small install, large corpus update. That is the ratio we watch for.