Enterprise AI Architecture Patterns
Most AI incidents are avoidable — they repeat because teams don't know the proven fix exists. 138 battle-tested patterns across 16 domains, mapped to APRA, EU AI Act, ISO 42001, NIST AI RMF, and GDPR.
Free · a few minutes · no sign-in required
Where does your org actually stand?
Three real patterns from the library — not a demo
Use it where you already work
Not another dashboard to remember to check. The library is queryable from your AI assistant and enforceable in your CI pipeline today.
Query patterns from your AI assistant
An MCP server, published on npm — no clone, no build. Your assistant can search the library, pull a pattern, or check a regulatory update mid-conversation.
{
"mcpServers": {
"aipatterns": {
"command": "npx",
"args": ["-y", "aipatterns-mcp-server"]
}
}
}Gate every PR on governance controls
Scans changed files for AI surface area, checks for the actual governance pattern in code — not just a filename — and posts a compliance report on the PR.
on:
pull_request:
types: [opened, synchronize]
jobs:
governance:
uses: eaapl/governance-action@v1
with:
api_key: ${{ secrets.EAAPL_API_KEY }}From AI risk to implementation
EAAPL is not a reading list. It is a structured path from identifying a risk to proving you've addressed it.
Assess your gaps — free, no sign-in
Answer questions about your AI systems and governance maturity. Get an instant risk score and your top 3 gaps, each mapped to the exact pattern that closes it.
Start assessment →Implement the patterns that matter
Not a generic checklist — every gap in your report links straight to the pattern for it, with the regulation it satisfies and the incidents it prevents.
Browse the library →Track adoption, prove governance
Create a free account, log which patterns are live, and export board-ready evidence — an APRA evidence package or a full audit trail.
Set up your tracker →Generate the architecture
Ready to build? Sign in and generate a starting architecture — diagram, components, and regulatory flags — for your specific use case.
Generate architecture →Where most teams begin
Three foundational patterns
Build these before anything else. They underpin every other pattern in the library.
AI Model Register
The AI Model Register is the foundational governance artefact for any enterprise…
Where every AI programme begins — a single source of truth for all models in production.
AI Gateway
The AI Gateway pattern establishes a centralised, enterprise-grade control plane through which…
The security perimeter every enterprise AI deployment requires before anything else goes live.
Enterprise Retrieval-Augmented Generation
Retrieval-Augmented Generation (RAG) is the foundational architecture pattern that grounds Large Language…
The grounding architecture that eliminates hallucinations and enables auditable AI responses.
Real incidents, preventable patterns
Every pattern in this library exists because an enterprise got it wrong first.
Air Canada chatbot fabricated a refund policy
Tribunal ordered compensation. Reputational damage and legal liability from a hallucinated AI response.
Samsung engineers leaked proprietary source code via ChatGPT
Confidential IP sent to an external LLM. No egress controls, no policy — data permanently outside the perimeter.
Agentic AI executed irreversible financial transactions autonomously
No human approval gate. Automated decisions with real-world consequences and no rollback path.
By Role
Find your reading path
Start with the patterns most relevant to your function.
CIO / CTO
Governance, risk posture, board reporting, and the investment case for AI.
Explore →Solution Architect
Reference architectures, pattern depth for proposals, and decisions you can justify to any stakeholder.
Explore →Enterprise Architect
Platform design, integration, and system composition across AI layers.
Explore →Security Architect
Zero-trust AI, input/output defence, prompt injection, and secrets management.
Explore →Consulting / Advisory
Structured discovery, gap analysis, and client evidence exports backed by real incident data.
Explore →AI Governance Lead
Model risk, policy enforcement, and regulatory compliance across multiple frameworks.
Explore →Platform Engineer
API gateways, cost control, observability, and CI/CD governance checks on every PR.
Explore →Compliance Officer
APRA CPS 230, CPS 234, EU AI Act, ISO 42001, NIST AI RMF, and GDPR — one library.
Explore →16 architecture domains
From foundational governance to agentic orchestration.
Agentic AI
10 patternsAutonomous agents, tool registries, sandboxing, and agent identity
Agentic Workflows
15 patternsReAct loops, sequential chains, parallel fan-out, routing, planning, and tool orchestration patterns
Regulatory Compliance
8 patternsAPRA CPS230/234, EU AI Act, ISO 42001, NIST AI RMF, and GDPR frameworks
Data Architecture
8 patternsTraining data governance, lineage, quality, and privacy-by-design
AI Governance
10 patternsModel lifecycle, risk classification, policy enforcement, and audit trails
Human-in-the-Loop
8 patternsApproval gateways, override patterns, confidence routing, and feedback loops
AI Integration
8 patternsService buses, circuit breakers, legacy augmentation, and batch processing
Knowledge Management
6 patternsKnowledge graphs, corpus management, vector databases, and semantic layers
Model Context Protocol
5 patternsMCP server design, gateway patterns, multi-server orchestration, auth/authz, and stateful session management
Model Management
8 patternsVersioning, rollback, canary releases, fine-tuning, and ensemble patterns
Multi-Agent Systems
6 patternsOrchestration, swarm patterns, debate agents, and handoff protocols
Observability & Monitoring
11 patternsTelemetry, drift detection, hallucination monitoring, and cost visibility
Platform Engineering
10 patternsAPI gateways, cost control, developer portals, caching, and multi-tenant platforms
Retrieval-Augmented Generation
10 patternsRetrieval architectures, vector search, grounding, and citation infrastructure
Reasoning Models
5 patternsExtended thinking gates, think-budget allocation, reasoning-then-act, cost-quality routing, and multi-step verification
AI Security
10 patternsZero-trust pipelines, input/output defence, secrets management, and access control
Compliance Mapping
Regulatory coverage
Every pattern is mapped to applicable frameworks. Demonstrate compliance readiness from day one.