Your organization models context implicitly.
The CMC Framework measures structural coherence across six structural levers to define the operating range for automated decision systems. If the decision density of a capability exceeds structural coherence in any single lever, deployment fails.
AI requires explicit ontology: entities, relationships, constraints in machine-readable form. Organizations run on implicit ontology: undocumented expertise, relationships understood but not formalised.
The gap between these isn't a spectrum. It's a category boundary.
Below the line: knowledge exists in conversations, experience, tribal memory. Above the line: knowledge exists in schemas, APIs, queryable structures.
EXPLICIT (L4–L5) means production infrastructure. The system enforces constraints. AI queries return valid results without human verification. This is “it works in production.”
PARTIAL (L2–L3) means pilot infrastructure. Documented, queryable, but unreliable. AI gets answers, often wrong. This is “works in the demo.”
IMPLICIT (L0–L1) means prototype infrastructure. In heads, emails, or unstructured files. AI cannot access it programmatically. This is “works in the pitch deck.”
AI cannot process what's below the line. Not “struggles with”. Cannot. There's no gradient between “in Maria's head” and “in a queryable schema.”
Gap ≥ 2 = the structure doesn't exist. Deployment is blocked.
This isn't novel theory. It's established computer science: automated reasoning requires explicit formal structures. The value isn't the insight. It's that most organizations and their consultants don't know it applies to them.
Per dimension:
Cn ≥ Required Cn
Deployment feasibility:
min(Cn − Required Cn) ≥ 0
Capacity (C) is measured per dimension per organisation. Required capacity is determined by the Autonomy Class of the capability being deployed. If any single dimension falls below its required level, deployment is structurally blocked. No dimension substitutes for another.
Each bar represents a decision density tier — from simple assistive automation to full system-level autonomy. The higher the tier, the more structural coherence is required.
The dashed line is your structural coherence — the weakest of six structural levers.
Capabilities below the threshold are within your operating range. Capabilities above it require structural investment before deployment — regardless of vendor quality or team ambition.
What determines that threshold? Six structural levers.
CMC measures stabilisation capacity across six structural dimensions. Each dimension answers one question about your infrastructure. Each is scored C0–C5. Integer levels only. If criteria for a level aren't fully met, the score is the level below.
A typical mid-market organization. Four dimensions below the category boundary. AI deployment: blocked.
Is context explicit or undocumented?
↳ from Knowledge Management (Polanyi, Nonaka & Takeuchi)
“When Maria's out, we're guessing.”
Level Progression:
How does context enter systems?
↳ from Business Process Management (Hammer, Davenport)
“That decision was made in a hallway conversation last year.”
Level Progression:
Is context organized for retrieval?
↳ from Information Architecture (DAMA-DMBOK, Rosenfeld & Morville)
“The information is in there somewhere.”
Level Progression:
Can AI reach the context?
↳ from Enterprise Architecture (TOGAF, Zachman)
“You need to ask IT for a report, and that takes two weeks.”
Level Progression:
Does context stay current?
↳ from Data Governance (DAMA, DataOps)
“That documentation is from 2019.”
Level Progression:
Do systems share context?
↳ from Systems Integration (Hohpe, Enterprise Integration Patterns)
“Our CRM and ERP don't talk to each other.”
Level Progression:
You cannot sustainably exceed your dependency by more than one level.
Formality L1 means Capture maxes at L2. Capture L2 means Structure maxes at L3. This isn't a suggestion. It's structural. Investing in Integration L4 while Accessibility is L1 is wasted spend.
Formality → Capture → Structure → Maintenance
↓
Accessibility → Integration
Each lever cannot sustainably exceed its dependency by more than one level.
Build sequence: always bottom-up.
This is why “just deploy AI” fails. It's roofing without foundations.
Five named failure patterns account for the majority of AI deployment failures: Islands (high Structure, low Integration), Amnesia (high Capture, low Maintenance), Locked Vault (high Formality, low Accessibility), Hero Dependency (low Formality across the board), and Baseline Low (all dimensions below C2). Each has a dimension signature that CMC diagnostics detect before procurement.
These are recognition aids. You should be able to identify your organization in one of them within 10 seconds.
"Our CRM is great. Our ERP is great. Getting them to share data is a six-month project."
Structure ≥ 3 AND Integration ≤ 1
"We document everything. Nobody trusts the documentation."
Capture ≥ 3 AND Maintenance ≤ 1
"It's all in SharePoint somewhere. Good luck finding it."
Formality ≥ 3 AND Accessibility ≤ 1
"When Maria's out, we're guessing."
Formality ≤ 1 AND average ≤ 1.5 (strict formality-driven bottleneck)
"Every pilot works in a sandbox. None survive production."
All dimensions ≤ 2 AND average < 2
Patterns derived from deployment failure analysis across research literature. Diagnostic confirms which pattern applies and identifies the build sequence to resolve it.
The CMC (Context Modelling Capability) Framework scores infrastructure feasibility for AI deployment across six dimensions: Formality, Capture, Structure, Accessibility, Maintenance, and Integration. Each dimension is scored 0-5, and gaps between what an AI capability requires and what an organization has determine deployment feasibility.
CMC compares an organization's infrastructure levels against the requirements of specific AI capabilities. Gaps of 2+ levels in any dimension indicate structural blocking — the organization lacks fundamental infrastructure that takes 12-24 months to build. This predicts deployment failure before procurement, not after.
Formality (how documented are processes), Capture (how knowledge is recorded), Structure (how standardized is data), Accessibility (how retrievable is information), Maintenance (how current is data), and Integration (how connected are systems). These follow two dependency branches from Formality: F → C → S → M and F → A → I. Each dimension cannot sustainably exceed its dependency by more than one level.
The CMC Framework maps 730 AI capabilities across 7 industries and 64 business functions. Each capability has infrastructure requirements scored across all 6 dimensions, creating 18,360 gap scenarios for feasibility analysis.
Researchers: Contact contact@contextcapability.com for dataset access and methodology documentation.
Infrastructure measurement precedes AI capital allocation.
See which AI capabilities your infrastructure can support, and which are blocked.
Check Deployability →30-minute conversation. We'll walk through your CMC House together.
Built on applied ontology: the computational foundation AI depends on. Methodology: Lean transformation. Applied ontology. FMEA process structure.