Seven Verticals, Seven Ontologies.
PH—013
Every vertical has its own regulatory reality, economic reality, and operating vocabulary. A generalist model approximates all three. An ontology codes them. This is the table, one filter at a time, across the seven verticals we currently ship.
I Why Vertical Matters More Than Model Choice
On thirty installs across seven verticals, the question “which model did you use” has come up maybe four times. The question “how do you handle the compliance surface in our industry” has come up on every single call.
Buyers are right to weight it that way. The model is a commodity. The compliance surface is not. The economic reality of the business is not. The vocabulary of the work is not. Those three things are what the ontology encodes, and they are what makes a recommendation correct or illegal, profitable or value-destroying, legible or noise.
II Public Innovation · The Civic Filter
The public-innovation ontology enforces four constraints unique to civic institutions. Data residency is non-negotiable. Certain recommendations cannot imply citizen-surveillance behavior, even when legally permitted. Narrative fit, which we added after the Port of San Antonio install, catches positioning that misrepresents the institution. Budget cycles are annual, not quarterly; the recommender respects the calendar.
III Wellness & Luxury · The Trust Filter
Trust is the product in wellness and luxury. The ontology enforces against any recommendation that short-cuts a brand promise for conversion gain. It also enforces against overly automated customer-facing communications in segments where high-touch is the expectation. Two wellness installs have surfaced recommendations that were economically sound and brand-destructive. Both were suppressed.
IV Home Services · The Same-Session Rule
Home services added the same-session-upselling rule after the Roadrunner install. A trade-service customer booking a septic service is in a transactional state that does not tolerate secondary offers. The filter suppresses recommendations that would violate this even when the gross-margin uplift is real.
V Hospitality · The Guest Boundary
The hospitality filter enforces what a concierge agent can commit to and what it must escalate. It also enforces privacy of guest preference data across properties in a group, even when sharing would improve personalization. The Concierge install hardened this filter.
VI Real Estate · The Comp Discipline
Real-estate recommendations must be anchored to real comparables. The filter prevents yield predictions based on small-sample comps. It also enforces a geographic heat threshold below which the recommender refuses to make a buy recommendation at all.
VII Sports & NIL · The Athlete Boundary
The sports and NIL filter enforces against any recommendation that would commit an athlete to an obligation beyond their eligibility window or their agent’s approval. It also enforces the separate handling of minor athletes. This filter is stricter than the regulatory baseline. The business case for the excess strictness is long-run: athletes who feel protected refer other athletes. Extraction does not scale.
VIII Consumer & Creative · The Attention Filter
The consumer and creative filter suppresses recommendations that compound attention at the cost of user wellbeing. We inherit this from the invisible design principles. The filter is weaker than we want it to be and we publish a refresh to it quarterly.
IX The Upkeep Cost
Maintaining seven ontologies in parallel is the most expensive thing we do. It takes three to four people full-time, distributed across the verticals. Each ontology receives a quarterly refresh in lockstep with the corpus. Each ontology receives ad-hoc updates whenever an install surfaces a new failure mode.
We have been asked, repeatedly, whether the ontologies could be consolidated. The answer is no. The filters overlap at the margin but disagree at the center. A single ontology that covers all seven verticals would be a softer filter than any of the seven individually. Generalization is the disease. Specificity is the treatment.
X The Next Three Verticals
Three verticals are under construction and will ship in the order below across the next four quarters. Healthcare commercialization. Education. Regulated manufacturing. Each will get its own volume in this research series when it ships.