Ontological governance
Automated classification systems drift over time — categories multiply, vocabulary fractures, and extracted data accumulates errors faster than they can be corrected. Ontological governance prevents this through two structural mechanisms: four throttled control ledgers that define classification vocabulary at intentionally slow update rates, and a human-verification loop that forces extracted identity fragments through human review before they are permanently written into the verified ledger. These two mechanisms are structurally separate but serve the same goal: preventing accumulated classification drift from undermining the integrity of long-lived institutional data. For a regulated operator, this means the platform's knowledge graph remains auditable and its identity records remain accurate without continuous manual curation.
The three-stage extraction pipeline
Data extraction across the platform is mechanically isolated into three services:
service-email(ingestion). Processes MIME payloads and deposits raw text and CSV files into the spool. No classification is applied at this stage.service-people(identity resolution). Scans the spool for human identity clusters and routes them to the verification surveyor before committing to the verified ledger.service-content(linguistic classification). Scans the spool for narrative knowledge and cross-references text against the four control ledgers.
The four control ledgers
service-content is governed by four CSV ledgers that update at
heavily throttled rates to preserve longitudinal data stability:
| Ledger | Minimum update interval | Governs |
|---|---|---|
| Archetypes | More than 24 months | The psychological and functional identity of the firm (for example, "The Fiduciary") |
| Chart of Accounts | 18–24 months; requires executive override | The structural and financial geometry of the operation (for example, "Compliance", "IT Support") |
| Domains | More than 12 months | Bilingual glossaries defining the macro-categories: Corporate (Finance), Projects (Real Estate), Documentation (Technology) |
| Themes | 3–8 months | The active frontline narratives (for example, "Co-Location Expansion") |
Update rates are intentionally asymmetric. The slowest ledgers (Archetypes, Chart of Accounts) capture what the firm fundamentally is; the fastest (Themes) capture what it is currently working on. Premature updates to the slower ledgers corrupt the longitudinal coherence of the data corpus.
The verification loop
service-people uses a human-in-the-loop verification step to
prevent automated extraction errors from entering the verified
ledger. The process is described in detail at
Verification Surveyor. In brief:
the system isolates unverified identity fragments for operator
review; the operator verifies each entity using their own personal
browser and off-network lookup; the verified result is then
committed to the ledger. The daily throughput limit ensures that
operator attention remains high-fidelity rather than habitual.