Davenport & Prusak (1998) define organizational knowledge as "a fluid mix of framed experience, values, contextual information, and expert insight" that "often becomes embedded not only in documents or repositories but also in organizational routines, processes, practices, and norms." They identify narrative as the primary transfer mechanism: "Knowledge communicated most effectively through a convincing narrative." The codification challenge is "more art than science." Technology is "the pipeline and storage system" but "does not create knowledge and cannot guarantee or even promote knowledge generation or knowledge sharing."

Daft & Lengel (1986) demonstrate that communication media vary in "information richness" — the capacity to change understanding within a time interval. Business plans and capability statements are rich precisely because they accommodate equivocality. For human readers, this is a feature. For computational governance, it means the same organizational knowledge is scattered across multiple documents in incompatible formats, each drifting from the others.

Studer, Benjamins & Fensel (1998) document the paradigm shift: "The modeling approach emphasizes the creation of computer models mimicking experts' problem-solving capabilities, countering the transfer approach's focus on simply encoding existing knowledge." Under the Modeling View, knowledge acquisition is model construction — building explicit models using reusable problem-solving methods and ontologies. The definition of ontology as "a formal, explicit specification of a shared conceptualization" (Studer et al., 1998) is one of the most cited in the field. Extracting organizational governance knowledge is a modeling activity: each answer to a structured question contributes to a model of organizational governance state, not a recording of domain knowledge.

The original 1987 IBM Systems Journal paper contained only three columns: Data (What), Function (How), and Network (Where). Who, When, and Why were added in the 1992 Sowa-Zachman paper "Extending and Formalizing the Framework." The full 6×6 matrix (six interrogatives × six reification transformations) makes an explicit completeness claim: "This matrix would necessarily constitute the total set of descriptive representations that are relevant for describing something...anything — in particular an enterprise." The completeness is justified by two independent classification systems deemed exhaustive. Critically, Zachman states: "The Zachman Framework IS NOT a methodology for creating the implementation...It clearly has no methodological implications...is an ontology — or the complete set of all the 'elements' that exist in the Enterprise." No extraction process, interview protocols, or discovery procedures are provided.

ArchiMate's business layer contains 13 element types: 4 active structure (Business Actor, Role, Collaboration, Interface), 5 behavior (Process, Function, Interaction, Event, Service), 3 passive structure (Object, Contract, Representation), and 1 composite (Product). The motivation layer adds elements including Stakeholder, Goal, Constraint, and Requirement. Across all layers, ArchiMate has approximately 57 element types, designed "to model the proverbial 80% of practical cases." Authority must be inferred through Role assignments. Decision requires Business Event with custom extension. Evidence has no native representation. Account (accountability state) has no native representation. This confirms a systematic gap: EA frameworks model organizational structure but not governance-specific dynamics.

CommonKADS's six models (Organization, Task, Agent, Knowledge, Communication, Design) include standardized context-level worksheets that work across domains: OM-1 (Problems, solutions, context), OM-2 (Organization focus area), OM-3 (Process breakdown), OM-4 (Knowledge assets), OM-5 (Feasibility), TM-1 (Task analysis), TM-2 (Knowledge item analysis), AM-1 (Agent model). These worksheets "act as checklist and information archive and should be used flexibly." At the knowledge model level, reusable templates exist for Classification, Assessment, Diagnosis, Monitoring, Synthesis, Configuration Design, Assignment, Planning, and Scheduling — "patterns of knowledge-intensive tasks." However, the Domain layer content within these templates requires domain-specific customization: "Modelling is a constructive activity, and there exists no single correct solution." The distinction matters: the structural method is standardized; the content requires domain expertise.

Klein et al.'s (1989) Critical Decision Method uses a four-sweep structure: (1) incident identification and unstructured recall, (2) timeline construction, (3) deepening probes with cognitive probes, (4) "what-if" queries. The probes target cues, goals, expectations, courses of action, experience/analogues, and information used — categories derived from Klein's Recognition-Primed Decision model. These probe categories are standardized across domains; the same RPD-derived questions apply to fireground commanders, surgeons, and software debuggers. What varies is the incident content, not the probe structure. Applied across "urban and wildland fireground commanders, tank platoon leaders, structural engineers, design engineers, paramedics, computer programmers" and many more domains. Inter-coder reliability for identifying decision points: 81-100%.

CTA extracts what experts actually do, perceive, and decide — inherently domain-specific because expert cognition varies by domain. A weather forecaster's cues differ from a critical care nurse's. But governance-level questions ("What types of work do you do?", "Who has authority to make this decision?", "What evidence would show this was done correctly?") extract structural knowledge that is the same regardless of domain. The audit tradition (F9) provides a century of empirical validation for this level distinction: auditors use standardized frameworks across every industry because governance structure — not domain content — is what they must understand.

ISA 315 (Revised 2019) requires understanding of: (1) industry, regulatory, and external factors, (2) nature of the entity (operations, ownership, governance, financing), (3) accounting policies, (4) entity objectives and strategies, (5) performance measurement, (6) applicable reporting framework, (7) the entity's system of internal control (all five COSO components), and (8) inherent risk factors. The COSO five-component model (control environment, risk assessment process, information system and communication, control activities, monitoring of controls) applies identically across all industries. The framework is "standardized in structure but scalable in application" — smaller entities satisfy requirements through informal processes. This is two-tier standardization: standardized framework, customized application.

Strategic-systems auditing (SSA) is "an evidence-driven, belief-based risk assessment process where the auditor develops an understanding of a company's business to form expectations about future financial statements." Three analytical stages: (1) strategic analysis (strategic risks from external forces, "the economic web"), (2) business process analysis (process risks — whether processes execute strategy), (3) business measurement (measurement risks — whether metrics capture reality). SSA treats the audit as a top-down, systems-level organizational modeling exercise — the auditor builds a model of organizational dynamics before examining any specific transaction. This is organizationally-scoped world modeling applied to governance assessment.

McCarthy (1982) formalized double-entry bookkeeping into three primitives: Economic Resources (goods, services, rights under control), Economic Events (occurrences that increment or decrement resources), and Economic Agents (persons accountable for or participating in events). The duality principle links paired events in exchange relationships — one increment, one decrement, representing economic reciprocity. Extended REA (Geerts & McCarthy, 2002) adds Commitments (promises of future events), Economic Contracts (commitment bundles), and Type Images (policy layer — budgets, pricing, bills of material). Even Extended REA does not natively model authority (who authorized a decision), evidence (what supported it), decision provenance (how choices were traced), or governance accountability (account of what happened and why).

Zachman's full 6×6 framework (Sowa & Zachman 1992; Zachman 2008) claims 36-cell completeness (classification, not governance). Osterwalder's BMC claims 9 building blocks sufficient for business model description (value creation, not governance). McCarthy's REA claims 3 concepts sufficient for accounting (economic exchange, not governance). Each tradition attempts to reduce organizational complexity to a sufficient set. Governance-specific constructs — authority, evidence, decision provenance, accountability state — are absent across all surveyed business ontologies. The gap is not in the formalization impulse but in the target domain.

Foster et al. (2007) define a lens as get : S → V (source to view, may discard information) and put : V × S → S (view plus original source back to source). Three laws: GetPut (no view change → no source change), PutGet (view change fully reflected in source), PutPut (sequential updates collapse). Well-behaved = GetPut + PutGet. The put function needs the original source because get typically discards information. For governance extraction, the "get" direction (narrative → structured representation) necessarily discards rhetorical structure, argumentative form, and stylistic choices. The "put" direction (structured representation → narrative) generates new rhetorical structure. GetPut is violated by design: extracting from a document and projecting back does not produce the original document. PutGet is approximately satisfied: projecting and re-extracting should recover the structured content.

QVT's limitations for non-bijective BX apply directly to governance knowledge extraction: narrative-to-structured-representation transformations are inherently non-bijective (many narratives can express the same governance content; one narrative can contain governance content for multiple structured elements). The resolution: asymmetric architecture where the structured representation is canonical and documents are derived views. This is a standard asymmetric lens pattern and aligns with DITA's single-source publishing paradigm (Rockley & Cooper, 2012) where source content is canonical and outputs are generated views.

In any content management system, the design decisions about what categories exist, what relationships hold, and what projection templates look like ARE the rhetorical acts. The decision that "organizational purpose" is a first-class governance concept — not a subsidiary of "business strategy" or "mission statement" — is a rhetorical act with consequences for every document projected from the structured representation. This connects to Studer et al.'s (1998) Modeling View: governance extraction is not neutral recording but model construction with built-in ontological commitments.

Enterprise Architecture. Provides organizational modeling ambition, multi-perspective decomposition, 57+ element types. Lacks extraction methodology; no governance-specific constructs (authority, evidence, decision, account absent).

Knowledge Engineering. Provides structured acquisition (templates, worksheets), Modeling View, competency questions. Targets knowledge-based systems, not governance; domain-specific content.

Cognitive Task Analysis. Provides expert methodology elicitation, structured interviews, progressive depth. Domain-specific probe design; no governance construct target.

Business Model Ontology. Provides formal element decomposition, sufficiency claims from minimal sets. Value creation or accounting scope; misses governance-specific constructs.

Bidirectional Transformation. Provides formal consistency framework (lenses, TGGs), correspondence models. Cannot achieve round-trip consistency for non-bijective transformations.

Structured Authoring. Provides single-source, multi-format publishing, content-presentation separation. Content is document topics, not governance state.

Audit Methodology. Provides standardized cross-domain organizational understanding, full governance dimension coverage. Output is risk assessment, not populated governance representation.

Noy & McGuinness establish competency questions as scope-defining instruments: questions that "a knowledge base based on the ontology should be able to answer." They serve as a "litmus test" for ontology coverage. The methodology (using competency questions to define scope) is standardizable across domains, but the content of the questions "necessarily varies by domain." The governance-level distinction (F8) resolves this: questions targeting governance structure rather than domain knowledge can be standardized because governance structure is invariant across organizational domains — as the audit tradition (F9) demonstrates empirically.