Intelligence Aggregates

Why This Exists

This document defines Intelligence Domain aggregate boundaries using Domain-Driven Design.

Aggregates protect consistency for intelligence orchestration, AI agent registration, reasoning sessions, memory, execution requests, guardrails, and audit records.

Owner

The owner is the Chief Product Officer and Enterprise Architect.

The Intelligence Domain owns these aggregate boundaries. Organization remains the root dependency for customer business context.

Business Value

Clear aggregate boundaries prevent AI orchestration from becoming ungoverned runtime behavior. They make intelligence work measurable, auditable, and safe.

Aggregate Catalogue

Aggregate	Purpose	Boundary
Intelligence Director	Coordinates intelligence work and routing.	Active orchestration decisions and assignments.
AI Agent Registry	Governs approved agents and Digital Professional runtime identities.	Agent definitions, statuses, allowed capabilities, guardrails.
Reasoning Session	Represents a bounded reasoning activity.	Inputs, context, reasoning steps, outputs, confidence, explainability.
AI Memory Record	Stores intelligence memory records and links to source context.	Memory entry, provenance, access scope, retention.
Execution Request	Represents a request to perform or advance executable work.	Requested action, approvals, guardrail status, execution result reference.
Guardrail Decision	Represents policy evaluation for human approval or automation.	Policy result, reason, approver, conditions.
Intelligence Audit Trail	Records immutable AI activity and decisions.	Audit entries grouped by correlation or session.

Aggregate Diagram

flowchart TD
    Director[Intelligence Director]
    Registry[AI Agent Registry]
    Session[Reasoning Session]
    Memory[AI Memory Record]
    Request[Execution Request]
    Guardrail[Guardrail Decision]
    Audit[Intelligence Audit Trail]
    Organization[OrganizationId Reference]

    Organization --> Session
    Director --> Registry
    Director --> Session
    Session --> Memory
    Session --> Request
    Request --> Guardrail
    Session --> Audit
    Request --> Audit
    Guardrail --> Audit

Core Invariants

Invariant	Description
Organization context is required	Customer-bound reasoning must include OrganizationId.
Agents must be registered	Digital Professionals or AI agents must be registered before orchestration.
Confidence must be explicit	AI outputs used for recommendations must include confidence scoring or an explicit exception.
Explainability must be captured	Recommendations and execution requests require explainability records.
Human approval must be enforced	Guardrail policies determine whether execution requires approval.
Audit is mandatory	Intelligence actions must create audit records.

Transaction Boundaries

Reasoning sessions should produce outputs and events without directly mutating Organization-owned facts. If Intelligence proposes changes to Organization, it should create a suggestion, event, or command handled through Organization APIs and policies.

Cross-Domain Reference Rule

Intelligence aggregates reference other domains by ID and event metadata. They do not own other domain aggregate state.