Table of Contents
- Purpose and Scope
- 1.1 Objectives
- 1.2 Non-Goals
- 1.3 Intended Audience
- Architectural Principles
- 2.1 Outcome-Oriented Design
- 2.2 Capability First, System Second
- 2.3 Deterministic by Default
- 2.4 AI as an Optional Executor
- 2.5 Loose Coupling and Replaceability
- 2.6 Governance Without Friction
- Key Definitions and Terminology
- 3.1 Capability
- 3.2 Capability Versioning
- 3.3 Context
- 3.4 Deterministic Execution
- 3.5 AI-Assisted Execution
- 3.6 MCP
- 3.7 Tool Invocation
- 3.8 Confidence and Provenance
- Architectural Overview
- 4.1 Logical Architecture
- 4.2 Physical Architecture
- 4.3 Trust and Security Boundaries
- 4.4 Data Flow Overview
- 4.5 Control Flow Overview
- Capability Model
- 5.1 Capability Identification
- 5.2 Capability Contracts
- 5.3 Capability Inputs and Outputs
- 5.4 Capability Lifecycle
- 5.5 Capability Ownership
- 5.6 Capability Maturity Levels
- Capability Invocation
- 6.1 Invocation Patterns
- 6.2 Context Propagation
- 6.3 Deterministic Invocation Path
- 6.4 AI-Assisted Invocation Path
- 6.5 Error Handling and Degradation
- Model Context Protocol (MCP)
- 7.1 Role of MCP in the Architecture
- 7.2 MCP Tool Definition Standards
- 7.3 MCP vs Direct API Invocation
- 7.4 MCP Discovery and Composition
- 7.5 MCP Security Considerations
- Digital Experience Platform (DXP)
- 8.1 DXP Responsibilities
- 8.2 Page Definition Model
- 8.3 Layouts and Slots
- 8.4 Capability Rendering
- 8.5 Navigation and Routing
- 8.6 Degraded Operation (AI Unavailable)
- User Interaction Model
- 9.1 Work-Oriented User Journeys
- 9.2 Work Aggregation and Queues
- 9.3 Filtering, Ranking, and Focus
- 9.4 Batch Actions
- 9.5 Deep-Dive Interaction
- Document Capabilities
- 10.1 Document Parsing Capability
- 10.2 Document Generation Capability
- 10.3 Template Governance
- 10.4 Audit and Provenance
- Integration Strategy
- 11.1 Integration Principles
- 11.2 Deterministic Integration Patterns
- 11.3 Semantic Integration Patterns
- 11.4 Legacy System Containment
- 11.5 Deferred Integration
- API Management and Gateway
- 12.1 Role of API Management
- 12.2 Capability Exposure
- 12.3 Versioning and Deprecation
- 12.4 Throttling and Quotas
- 12.5 Policy Enforcement
- Security and Identity
- 13.1 Authentication and Authorisation
- 13.2 Capability-Level Access Control
- 13.3 Data Sensitivity and Classification
- 13.4 Audit Logging
- 13.5 Trust Zones
- AI Governance and Controls
- 14.1 AI Usage Boundaries
- 14.2 Human-in-the-Loop Controls
- 14.3 Confidence Scoring
- 14.4 Explainability
- 14.5 Model Substitution and Rollback
- Performance and Cost Management
- 15.1 Latency Budgets
- 15.2 Deterministic vs AI Cost Profiles
- 15.3 Caching and Reuse
- 15.4 Scaling Characteristics
- Reliability and Resilience
- 16.1 Failure Modes
- 16.2 Graceful Degradation
- 16.3 Retries and Idempotency
- 16.4 Observability
- Development and Delivery
- 17.1 Capability Development Guidelines
- 17.2 Testing Strategy
- 17.3 Contract Validation
- 17.4 Continuous Delivery
- Operational Model
- 18.1 Monitoring and Metrics
- 18.2 Incident Management
- 18.3 Change Management
- 18.4 Operational Runbooks
- Governance and Compliance
- 19.1 Architectural Compliance
- 19.2 Capability Review Process
- 19.3 Audit Readiness
- 19.4 Regulatory Alignment
- Migration and Adoption Strategy
- 20.1 Incremental Adoption
- 20.2 Coexistence with Existing Platforms
- 20.3 Decommissioning Legacy Patterns
- 20.4 Success Metrics
- Use Cases and Scenarios
- 21.1 Document Generation
- 21.2 Progressive AI Adoption Pilot
- 21.3 Performance and Cost Impact
- Risks and Mitigations
- Open Decisions and Future Considerations
- Appendices
- A. Capability Catalogue Template
- B. MCP Tool Definition Template
- C. Page Definition Schema
- D. User Journeys and Work Orchestration
- E. Glossary