Relationship path evaluation
Checks whether a governed relationship path can support the question being asked.
Architecture Experience
Enterprise Reality architecture
Enterprise Reality is a constitutional architecture for enterprise cognition and governed decision-making. It is not a linear data pipeline; it is a progression of responsibility from governed publication through representation, reasoning, authorised action and assistance.
Runtime Progression
Constitutional responsibility progresses layer by layer
Enterprise Reality is not a linear data pipeline. Each layer receives governed inputs, performs a bounded responsibility and passes a more accountable representation forward.
Governed PublicationsStatus:Released
- Purpose
- Provide the approved body of public and governed programme knowledge.
- Primary responsibility
- Owns constitutional meaning, publication identity, lifecycle and review status.
- Key inputs
- Authored publications, governance records and approved metadata.
- Key outputs
- Stable public knowledge artefacts and publication evidence.
- Does not own
- Does not execute runtime reasoning, automation or operational decisions.
Metadata EngineStatus:Released
- Purpose
- Interpret publication metadata through the canonical metadata boundary.
- Primary responsibility
- Owns publication identity, status, relationships and metadata semantics.
- Key inputs
- Governed publication descriptors.
- Key outputs
- Canonical metadata models for repository and platform consumers.
- Does not own
- Does not own public narrative, graph reasoning or automation authority.
Enterprise CorpusStatus:In Development
- Purpose
- Organise governed publications into a discoverable enterprise body of knowledge.
- Primary responsibility
- Owns discovery, traversal, indexing, health and repository-level statistics.
- Key inputs
- Canonical metadata models and governed repository assets.
- Key outputs
- Corpus entries, repository views and health evidence.
- Does not own
- Does not parse publication metadata independently or decide truth.
Enterprise Reality GraphStatus:In Development
- Purpose
- Represent relationships between stable enterprise and publication identities.
- Primary responsibility
- Owns semantic relationship representation for governed knowledge.
- Key inputs
- Corpus entries and canonical relationship metadata.
- Key outputs
- Graph relationships suitable for reasoning and explanation.
- Does not own
- Does not replace source publications or become a second source of truth.
Enterprise Reasoning EngineStatus:In Development
- Purpose
- Resolve justified outcomes from governed relationships and evidence.
- Primary responsibility
- Owns path evaluation, evidence resolution, outcome resolution and explanations.
- Key inputs
- Enterprise Reality Graph relationships and governed evidence.
- Key outputs
- Reasoning outcomes, explanations and reviewable reports.
- Does not own
- Does not approve action, execute work or create policy authority.
Enterprise Automation EngineStatus:In Development
- Purpose
- Determine whether a justified reasoning outcome may become governed action.
- Primary responsibility
- Owns policy evaluation, eligibility, approval requirements and execution governance.
- Key inputs
- Reasoning reports, automation policy and explicit authority signals.
- Key outputs
- Automation decisions, action proposals, execution records and retry governance.
- Does not own
- Does not make reasoning true or treat proposal, mode or retry permission as authority.
Enterprise AI RuntimeStatus:Research
- Purpose
- Assist people in retrieving, explaining, summarising and navigating governed knowledge.
- Primary responsibility
- Owns assistive interaction patterns where they remain subordinate to constitutional authority.
- Key inputs
- Approved public knowledge, governed retrieval context and user prompts.
- Key outputs
- Assistance, summaries, recommendations and navigational support.
- Does not own
- Does not own truth, evidence resolution, policy, approval, decision or execution.
Ownership Model
Core ownership boundaries
The platform remains understandable because each component owns a narrow responsibility and explicitly does not own adjacent authority.
| Component | Owns | Does Not Own |
|---|---|---|
| Governed Repository / Publications | Approved publications, publication metadata, lifecycle status and governed meaning. | Runtime reasoning, operational execution, platform APIs or automation authority. |
| Metadata Engine | Canonical interpretation of publication metadata and publication identity. | Narrative content, repository traversal, graph reasoning or action decisions. |
| Enterprise Corpus | Discovery, traversal, indexing, repository health and corpus-level views. | Independent metadata parsing, semantic graph authority or reasoning outcomes. |
| Enterprise Reality Graph | Governed relationships between stable identities. | Source publication truth, policy authority, approval or execution. |
| Enterprise Reasoning Engine | Relationship path evaluation, evidence resolution, outcome resolution and explanation. | Automation permission, execution authority, approval or operational side effects. |
| Enterprise Automation Engine | Automation policy, eligibility, proposals, approval requirements, execution records and retry governance. | Reasoning truth, canonical evidence, human approval authority or source-system records. |
| Enterprise AI Runtime | Assistive retrieval, explanation, summarisation, navigation and recommendations. | Truth, evidence resolution, policy authority, approval, canonical decision or execution. |
Reasoning Pipeline
Reasoning determines what is justified
The reasoning pipeline evaluates governed relationships and evidence before producing an explanation. Validation observes. Statistics observe. Neither participates in reasoning.
- Enterprise Reality Graph
- Relationship Path Evaluation
- Evidence Resolution
- Outcome Resolution
- Explanation Generation
Evidence resolution
Resolves supporting, contrary or insufficient governed evidence without replacing absence with certainty.
Outcome and explanation
Produces a reviewable result and explanation rather than an unexplained assertion.
Automation Progression
Automation determines what governed action is permitted
Reasoning determines what is justified. Automation determines what governed action is permitted. Proposal is not authority. Execution mode is not authority. Approval and pre-authorisation are explicit. Decision, action proposal, execution authority and execution remain separate checkpoints.
- Reasoning Report
- Automation Policy
- Eligibility
- Automation Decision
- Action Proposal
- Approval or Pre-Authorisation
- Execution Authority
- Execution Request
- Execution
- Execution Record
- Idempotency and Retry Governance
Retry permission is not execution. The execution record is canonical history for what was attempted, under which authority, with which result.
Open-World Reasoning
Uncertainty remains explicit
Deterministic reasoning does not mean artificial certainty. Absence of evidence does not automatically mean false, and human review remains valid where evidence is insufficient.
Supporting governed evidence
SATISFIED
Explicit governed contrary evidence
NOT_SATISFIED
Insufficient governed evidence
INDETERMINATE
These semantics are not probabilistic AI inference. They preserve the distinction between known support, known contrary evidence and evidence that remains insufficient.
AI Boundary
AI assists. It does not establish constitutional authority.
AI may assist with retrieval, explanation, summarisation, navigation, recommendation and conversational assistance. It does not establish canonical truth, policy authority, approval authority or execution authority.
Permitted assistive roles
- Retrieval
- Explanation
- Summarisation
- Navigation
- Recommendation
- Conversational assistance
Prohibited constitutional ownership
- Truth
- Evidence resolution
- Policy authority
- Approval
- Execution authority
- Canonical decision
- Execution
The Assistant retrieves. It does not decide.
Constitutional Ladder
There is no direct constitutional path from Reality to AI
Assistance appears after governed representation, evidence, reasoning, decision, action and automation. It supports understanding; it does not bypass constitutional responsibility.
- TruthExists independently of systems and remains true whether or not it has been recorded.
- RealityThe enterprise state expressed through stable identities, events, relationships and obligations.
- RepresentationGoverned models that describe reality without claiming to replace it.
- EvidenceGoverned observations that support, contradict or leave an assertion unresolved.
- ReasoningDeterministic evaluation over governed relationships and evidence.
- DecisionA justified constitutional conclusion produced from reasoning and policy.
- ActionA governed response proposed as a consequence of a decision.
- AutomationCoordinated execution only after explicit authority exists.
- AssistanceAI retrieves, explains and recommends without owning truth, decision or authority.
Related Publications
Approved public basis
- ER-CON-002 — Enterprise Reality Website Constitution
- ER-WEB-001 — Enterprise Reality Website Specification
- ER-WEB-002 — Enterprise Reality Website Architecture Guide
- ER-WEB-003 — Enterprise Reality Website Design System
- ER-WEB-004 — Enterprise Reality Website Publication Guide