IBM's Azure Foundry Enterprise Kit
Bank of Canada · Enterprise AI WorkshopConfidential

Establishing Azure AI Foundry
as Your Enterprise AI Platform

This session covers the complete Microsoft AI Agent Platform stack, the enterprise governance and integration layer that accelerates enterprise adoption, and direct answers to Bank of Canada's specific architecture questions — all scoped to Canadian financial sector regulatory requirements.

Agent 365Foundry + Three-IQMCP FleetOSFI B-10Zero-Trust IdentityClean Core SAPDecision Ledger

What We Are Covering Today

This workshop is structured in two halves. We start with the Microsoft native platform — every GA and Preview feature relevant to large-scale enterprise AI deployment — then show the enterprise governance layer that accelerates adoption and closes the gaps Microsoft does not address natively. The second half takes each of Bank of Canada's open architecture questions and gives a layered, production-grade answer.

01
Microsoft Native Platform
23 GA / Preview features — Agent 365, Foundry, MCP governance stack, Three-IQ intelligence, Defender for AI, Purview.
02
Enterprise Kit — Governance Layer
What the kit adds on top: Decision Ledger, Certification Pipeline, 6 enterprise adapters, eval harness, banking policy packs, Bicep IaC.
03
AI Architecture at Scale
How do you prevent agent sprawl, run multi-agent workflows, and maintain a coherent governance topology across 50+ agents?
04
MCP Server Strategy
Private Tool Catalog, Foundry Toolbox versioning, APIM AI Gateway, 5 auth methods — full lifecycle governance for enterprise MCP fleets.
05
Identity & Principal Propagation
Three-layer Entra Agent Identity, FMI Path token exchange, zero-secret SAP/ServiceNow federation, Defender for AI runtime protection.
06
Data Flow & Auditability
Foundry IQ on OneLake, SAP Clean Core Bridge, Purview lineage graph, Decision Ledger hash-chain, OSFI B-10 evidence bundle.
07
Build Sequence
4-phase delivery from Week 1 zero-custom-code to Month 7 self-service platform — with concrete sprint deliverables.

What Microsoft Ships Out of the Box

The 2025–2026 Microsoft release cycle delivered a complete enterprise AI agent platform. The starting position is strong — do not build what Microsoft already provides. The Enterprise Kit configures and extends this platform; it does not replace it.

CapabilityMicrosoft FeatureStatusAnswers Bank Question
Agent 365Enterprise agent registry, lifecycle, AI Teammates, cross-cloud syncGAAI at Scale — sprawl prevention
Foundry Agent ServiceManaged agent runtime, tool invocations, multi-agent orchestrationGAAI at Scale — runtime
A2A Protocol v0.3Standardised agent-to-agent delegation with AgenticIdentity authGAAI at Scale — multi-agent
Foundry Continuous EvalAutomated quality gates, benchmarks, Red Teaming AgentGAAI at Scale — quality
Foundry MemoryCross-agent shared memory with TTL and access controlsPreviewAI at Scale — state
Foundry ToolboxVersioned, immutable MCP bundles with promotion gatesPreviewMCP Governance — versioning
Private Tool Catalog (API Center)Org-scoped MCP server registry with risk classificationPreviewMCP Governance — inventory
APIM AI Gateway for MCPRate limiting, geo-routing, semantic caching, correlation IDsPreviewMCP Governance — policy enforcement
MCP Auth (5 methods)API Key, OAuth Bearer, Managed Identity, mTLS, Entra SPNGAMCP Governance — auth
Entra Agent Identity — BlueprintApplication-level identity template with scoped permissionsGAIdentity — provisioning
Entra Agent Identity — BlueprintPrincipalDeployment-level service principal per agent classGAIdentity — deployment
Entra Agent Identity — per-instanceUnique identity per live agent instance with kill switchGAIdentity — instance isolation
FMI Path Token ExchangeFederated chain: Entra → SAP BTP → legacy system without secretsGAIdentity — SAP propagation
Defender for AIPrompt injection detection, adversarial input blocking at runtimeGAIdentity — runtime threat
Foundry IQOneLake-backed KB, permission-aware agentic retrieval, citationsGAData Flow — knowledge
Work IQM365 signal plane: Calendar, Mail, Teams, SharePoint, OneDrive, UserGAData Flow — M365
Fabric IQBI and analytics signal plane over Fabric semantic modelsGAData Flow — analytics
Purview Lineage GraphAgent-aware data lineage, column-level tracking to OneLakeGAData Flow — lineage
Purview DLP for AIPolicy-enforced content scanning on all agent tool invocationsGAData Flow — data boundary
Foundry GuardrailsContent safety filters, PII redaction, output schema validationGAData Flow — safety
Entra Admin Center — AI Agents viewCentralised agent inventory with last-active, permissions, kill switchGAAI at Scale — visibility
Azure Monitor / KQLStructured telemetry, latency dashboards, token-usage analyticsGAAll — observability
AI Foundry SDK (Python/TS)AIProjectClient, ToolboxClient, AgentClient scaffoldingGAAll — developer experience

What the Enterprise Kit Adds on Top of Microsoft

The Azure Foundry Enterprise Kit is not an alternative AI platform. It is a pre-built governance and integration layer that configures Microsoft's native stack for heavily regulated enterprise use. Six weeks of platform engineering is compressed to five days. Every module below maps to a control Microsoft does not provide natively.

Pillar 1 — Infrastructure
28 Bicep Modules
Subscription-scoped Azure CAF deployment. Identity, networking, Key Vault, Cosmos DB, AI Search, APIM AI Gateway, Content Safety, API Center, MCP hosting, hub-spoke networking.
Dev / Prod Profiles
Single-command deploy for dev sandbox or full production topology. All modules parameterised for multi-region.
Cost Management Module
Budget alerts, token-usage cost allocation, per-agent spend dashboards.
Pillar 2 — Agent Platform
6 Agent Templates
Knowledge Agent, Operations Agent, Policy Agent, Workflow Agent, Analyst Agent, Sample Invoice Agent — manifest-driven, Foundry Guardrails pre-wired, evaluation hooks included.
6 Enterprise Adapters
SAP (OData + JWT Bearer), ServiceNow (OAuth Passthrough), Salesforce, Workday, Microsoft Graph, Internal API Gateway. 350–500 lines each.
Agent Manifest Schema
Extended A365 manifest with: risk_tier, regulatory_scope, lob, duplicate_assessment, data_boundary, escalation_policy.
SDK Starters (Python / TS)
AIProjectClient + ToolboxClient + AgentClient scaffolding with telemetry, eval hooks, and managed identity pre-configured.
Pillar 3 — Governance
Decision Ledger
Hash-chained audit trail in Cosmos DB. Every agent decision, tool invocation, and escalation immutably recorded. Regulator-ready export.
Certification Pipeline
Bronze → Silver → Gold certification stages. Automated evidence collection, Red Teaming Agent integration, signed scorecard PDF.
Policy Graph
Cytoscape-based cross-agent compliance topology. Identifies policy violation paths across multi-agent workflows.
6 Policy Packs
foundry-baseline-v1, identity-strict-v1, toolbox-allowlist-strict-v1, data-boundary-v1, memory-retention-v1, escalation-policy-v1.
7 Eval Evaluators
Business intent, policy compliance, hallucination, groundedness, safety, PII leakage, latency. CI-wired with release gate thresholds.
RBAC Matrix
Pre-built role definitions for AI Operator, AI Auditor, AI Admin, Platform Engineer, Business Owner across all kit modules.
Pillar 4 — Business Value
Business Value Observability
Per-interaction ROI computation. 5 business scorecard templates (A–F grading). KPI dashboards with trend lines.
Drift Detection
8 drift detection rules. Automated remediation workflows triggered by threshold breach. Integrated with Azure Monitor.
Operations Guide
Day-2 runbook: incident response, capacity planning, model swap procedures, rollback playbooks.
Positioning: Three Layers
Layer 1 — Microsoft Native (GA/Preview)

Agent 365, Foundry, Toolbox, Private Tool Catalog, APIM AI Gateway, Entra Agent Identity, Defender for AI, Foundry IQ, Work IQ, Fabric IQ, Purview.

Layer 2 — Enterprise Kit (Configuration & Governance)

Decision Ledger, Certification Pipeline, Policy Graph, enterprise adapters, eval harness, banking policy packs, Bicep IaC, business value dashboards.

Layer 3 — Banking Extensions (To Build Together)

OSFI B-10 evidence bundle export, bank-specific MCP governance YAML, identity propagation registry, SAP Clean Core Bridge, Agent Portfolio Registry manifest extension.

AI Architecture at Scale: Governing 50+ Agents

Bank of Canada's Question

“How do we prevent agent sprawl? How do we ensure every agent across business units is visible, governed, and compliant — without creating a governance bottleneck?”

Microsoft — GA May 2026Agent 365: Enterprise Agent Registry

Agent 365 is the sprawl prevention platform. Every agent — deployed anywhere in the Microsoft ecosystem — registers here automatically. Four-type registry: Domain Agents, Cross-Domain Agents, Orchestrator Agents, Specialist Agents.

  • AI Teammates integration: agents appear alongside human colleagues in M365 org chart, with last-active timestamp and assigned owner
  • Cross-cloud registry sync: Azure + AWS Bedrock + GCP Vertex agents unified in one control plane
  • A365 CLI gate: a365 register --manifest agent.yaml --env prod — enforced in CI, no unregistered deployments
  • Kill switch: a365 disable-agent --id <agent-id> --reason "policy-violation"
Microsoft — GAA2A Protocol v0.3: Multi-Agent Orchestration

The A2A Protocol defines how agents delegate to each other safely. Orchestrator agents discover sub-agent capabilities via Agent Cards (JSON-LD), then delegate with AgenticIdentity-scoped tokens — the sub-agent can only do what the orchestrator explicitly delegates.

  • Skill discovery: GET /.well-known/agent-card — every registered agent exposes its capabilities
  • Delegation chain: tokens carry the full orchestration path for audit purposes
  • Foundry Memory (Preview): shared working memory across a multi-agent workflow session with TTL and conflict resolution
Enterprise KitEnterprise Kit: Policy Graph + Certification Pipeline

The Enterprise kit adds the governance controls Microsoft does not provide: a cross-agent compliance topology that visualises violation paths, and an automated certification pipeline that agents must pass before production.

  • Policy Graph: Cytoscape topology of all registered agents, with edges representing tool-sharing relationships and policy dependencies. Identifies compliance violation paths before they reach production.
  • Certification Pipeline: Bronze (basic safety) → Silver (policy compliance + eval gates) → Gold (OSFI evidence bundle + Red Teaming passed). No Gold → no production deployment.
  • Agent Manifest Extension: risk_tier, regulatory_scope, lob, duplicate_assessment — banking-specific manifest fields appended to A365 base schema.
  • Scorecard Compiler: Automated PDF scorecard per agent per release, signed by certification pipeline, suitable for board-level AI governance reporting.

MCP Server Strategy: Fleet Governance at Enterprise Scale

Bank of Canada's Question

“We will have hundreds of MCP servers across business units. How do we version them, control access, enforce policy, and prevent rogue tool servers from being connected to production agents?”

Microsoft — Preview5-Feature MCP Governance Stack
1
Private Tool Catalog (API Center)
Organisation-scoped registry of approved MCP servers. Each entry carries risk classification (low / medium / high / restricted), owner, and data-boundary tags. No unregistered server can be connected to a production agent.
2
Foundry Toolbox
Versioned, immutable MCP bundles. Promotion pipeline: dev → staging → prod with approval gate. Rollback: one command. Dependency pinning prevents supply chain drift.
3
APIM AI Gateway for MCP
Every MCP invocation passes through APIM. Semantic caching (identical tool calls use cached response), per-server token limits, geo-routing (Canadian data residency), correlation IDs for full trace across agent → MCP → backend.
4
5 MCP Authentication Methods
API Key, OAuth 2.0 Bearer, Managed Identity (zero-secret), mTLS (mutual TLS), Entra Service Principal. Each method carries a different risk tier in the Private Tool Catalog.
5
Toolbox Allowlist Policy
Per-agent toolbox allowlist enforced at runtime. An agent can only call tools explicitly listed in its manifest. Enterprise policy pack: toolbox-allowlist-strict-v1.
Enterprise KitEnterprise Kit: 6 Pre-Built Enterprise Adapters

The Enterprise Kit provides production-ready MCP adapters for the systems Bank of Canada already uses. No adapter code to write from scratch.

SAP Adapter
OData V4 + JWT Bearer (FMI Path). Supports delta tokens, S/4HANA operations, AP/AR/GL read. 200+ lines.
ServiceNow Adapter
OAuth Passthrough, ITIL workflow tools, incident/change management, CMDB read/write. 350+ lines.
Microsoft Graph Adapter
Calendar, Mail, Teams, SharePoint, OneDrive, User tools via Managed Identity. Work IQ integration.
Salesforce Adapter
OAuth 2.0, Salesforce REST API, SOQL query tools, Apex trigger hooks.
Workday Adapter
REST API, HR data (employee, org structure, benefits), Workday RaaS reports.
Internal API Gateway
Generic adapter template for internal REST APIs. Auth injection, rate limiting, retry logic pre-wired.
Banking Extension — To Buildbank-mcp-governance.yaml

A bank-specific MCP governance overlay extending the Private Tool Catalog with OSFI-aligned risk classification, Canadian data residency enforcement, and automated audit export.

# bank-mcp-governance.yaml
tool_risk_classification:
  tiers:
    tier_1_low:      [graph-calendar, graph-mail, foundry-iq-search]
    tier_2_medium:   [servicenow-incident, workday-hr-read]
    tier_3_high:     [sap-finance-write, servicenow-change-approve]
    tier_4_restricted: [core-banking-ledger, wire-transfer-api]

data_boundary:
  canadian_residency_required: true
  allowed_regions: [canadacentral, canadaeast]
  restricted_data_types: [PII_SIN, SWIFT_codes, OSFI_regulated_data]

audit:
  export_format: osfi_b10_signed_bundle
  retention_days: 2555   # 7 years per OSFI B-10 guideline
  hash_algorithm: SHA-256-chain

Identity & Principal Propagation: Zero-Secret Enterprise Auth

Bank of Canada's Question

“How does the agent identity propagate through to SAP, ServiceNow, and our internal systems without creating shared service accounts or storing credentials? And how do we prove to auditors which human authorised a given AI action?”

Microsoft — GAThree-Layer Entra Agent Identity Architecture
Blueprint
Application-level
One Blueprint per agent class (e.g., "Finance-Analyst-Agent"). Defines the permission template — which APIs, which scopes, which data classifications. Created once by the AI Platform team.
BlueprintPrincipal
Deployment-level
One BlueprintPrincipal per deployment environment per line of business. "Finance-Analyst-Agent :: FX-Desk :: Prod". Inherits Blueprint permissions, adds environment-specific RBAC.
Per-Instance Identity
Runtime
One unique identity per live agent instance. Isolated credentials, isolated session. Compromise of one instance does not affect siblings. Kill switch: disable_blueprint() terminates the entire lineage.
Microsoft — GAFMI Path Token Exchange: Zero-Secret SAP & ServiceNow

The FMI Path (Federated Managed Identity) eliminates shared service accounts entirely. The agent's Entra identity becomes a Federated Identity Credential that exchanges for a SAP BTP token without storing any credentials anywhere.

Entra Agent Token
→ APIM AI Gateway (correlation ID attached)
→ FMI Path Exchange
→ SAP BTP OAuth
→ SAP S/4HANA OData API

Each hop carries the correlation ID. The Decision Ledger records: originating human session (if any), agent instance ID, FMI exchange timestamp, SAP action taken. This is the complete delegation chain auditors require.

Enterprise KitEnterprise Kit: Identity Propagation Registry + Decision Ledger
# identity-propagation-registry.yaml (Enterprise Kit extension)
entra_group_to_system_roles:
  - entra_group: "BoC-FX-Traders"
    sap_auth_object: "F_BKPF_BES"
    servicenow_role: "financial_operations_reader"
    allowed_data_tiers: [internal, confidential]

  - entra_group: "BoC-AI-Platform-Admins"
    sap_auth_object: "S_TABU_DIS"
    servicenow_role: "admin"
    allowed_data_tiers: [internal, confidential, restricted]
  • Decision Ledger: Hash-chained Cosmos DB log. Every SAP write, every ServiceNow change, every agent decision — immutably recorded with originating identity chain.
  • OSFI B-10 Audit Export: Certification Pipeline packages the Decision Ledger, Entra audit logs, and Purview lineage into a signed bundle. Directly submittable to OSFI examiners.
  • Defender for AI (Microsoft GA): Prompt injection detection at runtime. If an adversarial input attempts to override the agent's identity or escalate permissions, it is blocked and logged.

Data Flow & Clean Core: From SAP to Foundry IQ Without Data Copies

Bank of Canada's Question

“How do agents access our SAP financial data, regulatory data, and internal knowledge bases without creating data copies that breach our clean-core SAP policy or Canadian data residency requirements? And how do we trace every data access for auditability?”

Microsoft — GAFoundry IQ: Permission-Aware Agentic Retrieval on OneLake

Foundry IQ is the knowledge layer. Data lives in OneLake (Canada Central). Agents query Foundry IQ — which enforces the querying agent's permissions before returning any content. No data copies leave the sovereignty boundary.

  • Multi-source KB: SharePoint, SAP extracts, regulatory documents, internal policies — all indexed into Foundry IQ with column-level Purview classification
  • Agentic retrieval: semantic search with permission enforcement. An agent with finance-read scope cannot retrieve HR data even if it asks for it.
  • Citations: every response includes source attribution with Purview lineage record. Auditors can trace any agent output back to its source document.
  • Work IQ: M365 signal plane — Calendar, Mail, Teams, SharePoint, User. Foundry IQ + Work IQ = complete organisational knowledge.
Banking Extension — SAP Clean Core BridgeSAP → OneLake Bronze/Silver/Gold Pipeline

The SAP Clean Core Bridge extracts SAP financial data into OneLake using a CDC (Change Data Capture) pattern — no direct SAP queries from agents, no data copies outside Azure Canada region.

SAP S/4HANA
OData V4 delta token extraction Event Mesh → Event Hubs
Azure Data Factory
Incremental load pipeline Canadacentral only
OneLake Bronze
Raw financial records Purview column classification
OneLake Silver
Validated, deduplicated CDM format
OneLake Gold
Aggregated, BI-ready Fabric IQ semantic model
Foundry IQ
Permission-aware retrieval Agent-facing KB
  • SAP clean-core preserved: no custom SAP extensions. OData V4 is SAP's standard API surface.
  • Canadian data residency: all ADF pipelines, OneLake, and Foundry IQ scoped to canadacentral / canadaeast.
  • Purview lineage: every row in OneLake carries a lineage record back to its SAP source table and column.
Enterprise KitEnterprise Kit: Decision Ledger + OSFI B-10 Audit Bundle
OSFI B-10 §4.3 — Model Risk
Certification scorecard PDF (Bronze → Gold), eval pack results, Red Teaming Agent pass report
OSFI B-10 §4.4 — Third-Party Risk
MCP adapter audit logs, APIM gateway invocation records, Foundry Toolbox version manifest
OSFI B-10 §5.1 — Data Governance
Purview lineage graph export, OneLake classification report, data boundary policy enforcement log
FINTRAC — Transaction Monitoring
Decision Ledger hash-chain export (all financial tool invocations), SAP OData access log
PIPEDA — Personal Information
PII redaction log (Foundry Guardrails), Purview DLP policy enforcement record, consent management trail
Canadian Data Residency
Bicep deployment proof (canadacentral only), Azure Policy compliance report, ADF pipeline region lock

From Zero to Production Platform: 7-Month Sequence

The build sequence is designed around zero custom infrastructure code in the first sprint. Every phase builds on the previous with no throwaway work.

Phase 1Week 1–2
Zero custom code
Platform Foundation
  • Deploy Enterprise Kit Bicep IaC: subscription-scoped Azure CAF baseline
  • Provision Foundry project, Agent 365 tenant, APIM AI Gateway, API Center
  • Configure Entra Agent Identity: 1 Blueprint per agent class
  • Register 3 pilot agents in Agent 365 with extended banking manifest
  • Activate foundry-baseline-v1 and data-boundary-v1 policy packs
  • Wire Foundry Continuous Eval + Red Teaming Agent
Phase 2Month 2–3
First custom code
Banking Extensions
  • Customise bank-mcp-governance.yaml for BoC risk classification
  • Configure identity-propagation-registry.yaml (Entra → SAP auth objects)
  • Deploy SAP Adapter + ServiceNow Adapter into Private Tool Catalog
  • Implement Foundry IQ multi-source KB: internal policies, regulatory docs
  • Integrate Decision Ledger with SAP Adapter (every write is ledgered)
  • Run first Bronze Certification for 3 pilot agents
Phase 3Month 4–6
Full data pipeline
Clean Core Bridge + Certification
  • Build SAP → OneLake Bronze/Silver/Gold ADF pipeline
  • Configure Purview column classification on all OneLake tables
  • Connect Fabric IQ semantic models to OneLake Gold
  • Achieve Silver Certification for all pilot agents (eval gates passing)
  • Wire OSFI B-10 audit bundle export into Certification Pipeline
  • Load test at 5x target throughput (k6 load test kit included)
Phase 4Month 7+
Scale & govern
Self-Service Platform
  • Enable self-service agent registration via A365 CLI gate in DevOps pipeline
  • Publish Private Tool Catalog to all business units (with tier-based access)
  • Achieve Gold Certification for production agents
  • Activate business value dashboards (ROI, scorecard A–F per agent)
  • Establish quarterly OSFI B-10 evidence bundle export cadence
  • Expand to full enterprise: 50+ agents, 10+ MCP server fleet
Executive Summary

We are not building a custom AI platform on top of Azure. We are configuring Microsoft's enterprise AI platform — Agent 365 for governance, Foundry for runtime, MCP for tool connectivity, and the Three-IQ intelligence layer for knowledge and analytics — with the enterprise governance and banking-specific integration layer on top.

Bank of Canada gets a production-grade, OSFI-compliant AI agent platform in 7 months, with zero throwaway infrastructure work, using tools that are already procured through the Microsoft enterprise agreement.