Skip to content

Intelligence Business Rules

Why This Exists

This document defines business rules for the Intelligence Domain.

These rules ensure AI orchestration and reasoning remain governed, explainable, auditable, and aligned to business ownership.

Owner

The owner is the Chief Product Officer and Enterprise Architect.

Business Value

Business rules protect customer trust and prevent ungoverned AI behavior.

Rule Catalogue

Rule ID Area Rule
INT-RULE-001 Organization context Customer-bound reasoning must include OrganizationId.
INT-RULE-002 Domain ownership Intelligence must not mutate Organization-owned facts directly.
INT-RULE-003 Agent registry AI agents must be registered before use.
INT-RULE-004 Digital Professionals Digital Professional runtime must map to an approved Digital Professional.
INT-RULE-005 Reasoning Reasoning sessions must record objective, context, output, confidence, and explanation.
INT-RULE-006 Confidence Customer-facing AI recommendations require confidence scoring or explicit exception.
INT-RULE-007 Explainability Recommendations and execution requests require explainability records.
INT-RULE-008 Human approval Guardrail policy determines whether approval is required before execution.
INT-RULE-009 Audit Intelligence actions must create audit records.
INT-RULE-010 Memory AI Memory records must include provenance and scope.
INT-RULE-011 Event Mesh Intelligence must publish meaningful domain events for cross-domain reactions.
INT-RULE-012 Execution Intelligence may request execution but must not bypass execution guardrails.
INT-RULE-013 AI suggestions AI-enriched suggestions require review before becoming canonical domain facts.
INT-RULE-014 Evidence Reasoning must preserve evidence references where available.

Human Approval Rule

Human approval is required when AI output may materially affect customer commitments, submissions, compliance posture, contract obligations, funding decisions, or external communications.

Explainability Rule

Explainability must describe:

  • Evidence used.
  • Reasoning summary.
  • Confidence level.
  • Assumptions.
  • Limitations.
  • Required human action.

Audit Rule

Audit logging must be append-only in principle and must support correlation across sessions, events, approvals, and execution requests.