Enterprise Architecture · AI-Modernized Reference Model

TIERS & LAYERS — RESPONSIBILITIES

SEPARATION OF CONCERNS ENFORCED · AI ERA EXTENSIONS · 2026
Presentation / Product (Original)
Service Tier (Original)
Recordkeeping (Original)
Security (Original)
Infrastructure / Shared Lib
NEW — AI-Era Addition
Consumer
Producer
Presentation Tier
Device · Channel · UX Rendering
Product Tier
Interaction · Auth · Coarse AuthZ · UI Logging
AI / Intelligence Tier NEW
Models · Agents · Inference · RAG
Service Tier
Integration · Routing · Protocol
Recordkeeping Tier
SoR · Business Logic · Fine-Grained AuthZ · Txn Logging
Logic Sub-Layer
Presentation
Device-Specific Rendering
Globalization / Localization
Accessibility (A11y)
Conversational / Chat UI NEW
Copilot & AI Widget Rendering NEW
AI-Generated Content Display NEW
User Interaction & Appearance
User Interaction & Appearance
Producer Interaction (UI)
Globalization (UI)
Natural Language Input Handling NEW
AI Response Streaming / Display NEW
AI Logic
LLM / Foundation Model Serving NEW
Inference Engine NEW
RAG Pipeline NEW
Agent Orchestration NEW
Prompt Mgmt & Versioning NEW
Fine-Tuning Pipeline NEW
Model Registry NEW
Service Routing
Service Access
Recordkeeper Interaction
Protocol Translation
Reliable Messaging / File Transfer
AI API Gateway NEW
Event Streaming (Kafka/Kinesis) NEW
Model Endpoint Routing NEW
Business Logic & SoR Access
Access to Information
Transaction Processing
Online / Batch Processing
Data Integrity Enforcement
Recordkeeper Interaction Logic
Reporting & Analytics Interaction
AI-Enriched Record Processing NEW
Feature Store Access NEW
Real-Time Streaming Ingestion NEW
Training Data Lineage & SoR NEW
⚑ Business logic and data validation rules reside exclusively in the Recordkeeping Tier. The AI Tier orchestrates and reasons — it does not own business rules or data.
Auth — N/A
No auth responsibility
Authentication & Coarse-Grained AuthZ
User Authentication (who are you?)
Coarse-Grained AuthZ (product / feature access)
Session Establishment
AI Session Identity & Trust Token NEW
Copilot Access Control (coarse) NEW
AI Model Security
Model-Level Access Control NEW
Prompt Injection Detection NEW
Output Filtering / Guardrails NEW
Auth — Pass-Through Only
Token / Credential Forwarding
Service-to-Service Auth
AI Service Auth Token Relay NEW
Fine-Grained / Data-Level AuthZ
Fine-Grained AuthZ (row / field level)
Data-Level Access Enforcement
Producer Identity Verification
AI Model Output Data AuthZ NEW
Feature Store Access Control NEW
⚑ Security split: Product Tier = authentication + coarse-grained AuthZ. Recordkeeping Tier = fine-grained / data-level AuthZ. Service Tier relays credentials only. AI Tier enforces model-level guardrails.
Logging — N/A
No logging responsibility
User Interaction Logging
User Activity Logging (clicks, navigation)
Session Event Logging
UI Error Logging
AI Copilot Interaction Logging NEW
Prompt Input / Response Logging NEW
AI Operational Logging
Model Inference Logging NEW
Model Drift & Performance Monitoring NEW
Hallucination / Quality Flagging NEW
Experiment Tracking NEW
Integration Logging
Service Call Logging
Integration Error Logging
AI API Call Logging NEW
Transaction / Audit Logging
Transaction Audit Logging
Batch Process Logging
Data Access Audit Trail
AI-Driven Transaction Audit NEW
Training Data Access Audit NEW
⚑ Logging split: Product Tier = user interaction logs. Recordkeeping Tier = transaction / audit logs. AI Tier = model operational logs. Systems-level logging belongs in the Platform Layer only.
Config — N/A
No config responsibility
Consumer-Side Configuration
Consumer UI Configuration
Feature Flags (UX)
AI Preference Profiles NEW
Copilot Personalization Config NEW
AI Model Configuration
Model Hyperparameter Config NEW
Prompt Template Management NEW
RAG Retrieval Config NEW
Service Config
Integration Endpoint Config
Message Routing Config
Producer-Side Configuration
Producer / Recordkeeper Config
Business Rule Configuration
Feature Store Config NEW
Data Schema & Lineage Config NEW
N/A
No reporting
Consumer-Facing Reports
Consumer-Facing Dashboards
AI-Generated Consumer Insights NEW
Personalized Report Rendering NEW
AI Performance Reporting
Model Performance Dashboards NEW
A/B Model Testing Reports NEW
N/A
No reporting
Authoritative Reporting & Analytics (SoR)
Reporting & Analytics
Data Warehouse / Mart
Data Lakehouse (AI-era SoR) NEW
Predictive / AI-Augmented Analytics NEW
Application Layer — Data Sub-Layer
Consumer — Channel / Content / Session Data
Channel Data
Content Data
Session Data
Interaction / Clickstream Data NEW
Copilot Context Data NEW
Producer — Authoritative Systems of Record
Recordkeeping Data (SoR)
Data Warehouse / Mart
Data Lakehouse (AI-era SoR) NEW
Feature Store (ML SoR) NEW
Vector Database NEW
Embedding Store NEW
⚑ All Producer-side data stores are authoritative Systems of Record. Feature Store, Vector DB, and Data Lakehouse are AI-era SoR extensions — no other tier maintains its own copy of business data.
Infrastructure Layer
Domain-Agnostic Shared Libraries — consumed by every tier's applications via the Java I-Layer (JIL) pattern. No business logic resides here; these are technical utilities shared uniformly across all tiers.
Common Logging Libraries
Shared Security / Encryption Libs
Session Management Libraries
Globalization / i18n Framework
Configuration Service Client
Common Auth Token Validation
Connection Pooling Libraries
Service Client Stubs
Container Orchestration (Kubernetes)
Service Mesh (Istio)
MLOps / CI-CD Shared Pipeline NEW
AI Observability Libraries NEW
Shared Prompt Injection Defense Lib NEW
AI Ethics & Bias Evaluation Framework NEW
Model Governance Policy Engine NEW
Data Privacy / GDPR Compliance Libs NEW
⚑ Shared technical libraries (JIL pattern) consumed by all tiers. No business logic belongs here.
Environment Layer
Vendor-Provided Platform Services — App Servers · RDBMS · Message Brokers · NoSQL Stores · Identity Providers · API Management
Application Servers
Relational Databases
Message Brokers
Identity Providers (IdP)
GPU / TPU Compute Clusters NEW
Vector DB Platform (Pinecone / Weaviate) NEW
ML Platform (MLflow / Kubeflow) NEW
Platform Layer
Hardware & OS — Physical/virtual hardware · Operating Systems · Network infrastructure · System Management · Cloud / Hybrid / On-Prem IaaS
Hardware
Operating Systems
Network & System Management
Performance / Error / Event Logging (Systems)
Cloud AI Services (SageMaker / Azure OpenAI / Vertex) NEW
GPU Infrastructure Monitoring NEW
⚑ Systems-level logging (Performance, Error/Trace, Event) belongs here — not in application tiers.
TOGAF-ALIGNED · SAFe-COMPATIBLE · SEPARATION OF CONCERNS ENFORCED · AI EXTENSIONS · SEE COMPANION: TIERS & LAYERS KEY CONCEPTS
■ RED = AI-Era  |  ■ YELLOW = Presentation/Product  |  ■ BLUE = Service  |  ■ GREEN = Recordkeeping  |  ■ TEAL = Security  |  ■ DARK GREEN = Infrastructure / Shared Lib
© 2026 Mark Murray · All Rights Reserved  |  Release 1.0  |  March 2026