Skip to content

Framework Overview

Current version: v1.6.2 (May 2026)

See What's New for the latest release notes and the full version history.

The FSI Agent Governance Framework provides comprehensive governance guidance for Microsoft 365 AI agents (Microsoft Copilot Studio, Agent Builder) in US financial services organizations.


Purpose

This Framework layer establishes the foundational governance principles, organizational structure, and regulatory context for AI agent deployment. Content here is designed for:

  • Executives and Board Members — Strategic oversight and risk appetite decisions
  • Compliance Officers — Regulatory alignment and examination readiness
  • AI Governance Committees — Policy decisions and approval workflows
  • Auditors — Framework structure and control objectives

Three-Layer Documentation Architecture

The FSI Agent Governance Framework uses a three-layer documentation model to separate stable governance principles from frequently-updated implementation procedures:

Layer Content Update Frequency Files
1. Framework (this layer) Governance principles, zones, lifecycle, regulatory context 1-2x per year 15 pages (this Overview + 14 documents)
2. Control Catalog 79 technical control specifications across 4 pillars Quarterly 79 control files
3. Playbooks Step-by-step implementation procedures Continuous (as Microsoft portals change) 316 playbook files (4 per control)

This separation ensures governance stability while allowing rapid updates to implementation guidance as Microsoft 365 and Power Platform evolve.


Framework Concepts at a Glance

Five interlocking concepts structure this framework. Admins new to FSI AgentGov need Zone, Pillar, and Control to get started — Pattern and Capability Driver belong to the Advanced & Scaling track and can be deferred until the core model is working.

Concept One-line definition Where defined
Zone Risk classification for an agent: Personal (Zone 1), Team (Zone 2), or Enterprise (Zone 3). Determines which controls apply, approval requirements, and retention periods. Zones and Tiers
Pillar Control family — Security (29 controls), Management (27), Reporting (14), SharePoint (9). Answers "what category is this control?" — not to be confused with Capability Drivers. Governance Fundamentals
Control A single actionable governance unit — 79 total, each with Baseline / Recommended / Regulated implementation levels. The primary unit of day-to-day governance work. Control Catalog
Pattern Deployment shape (one of six CAPE Frontier Transformation Patterns). Useful for translating CIO conversations into zone and control requirements. Transformation Patterns
Capability Driver Organizational readiness dimension (one of five). Measured on a separate 100–500 CAPE scale — distinct from the 0–4 control implementation level. Capability Drivers

⚡ = Advanced & Scaling track. M365 admins implementing controls can skip Pattern and Capability Driver entirely. Navigate to the Adoption Roadmap or Control Catalog to begin implementation.


Framework Components

Core Documents

Document Purpose Audience
Executive Summary Board-level overview of AI agent risks and governance C-suite, Board
Governance Fundamentals Core framework concepts, maturity scale definition, and structure All stakeholders
Zones and Tiers Three-zone governance model Governance committees
Agent Lifecycle Lifecycle phases and governance requirements Compliance, Operations
Regulatory Framework US regulatory requirements and control mappings Compliance, Legal
Operating Model RACI, roles, governance structure All stakeholders
Governance Cadence Review schedules and audit readiness Compliance, Audit
Adoption Roadmap Phased implementation guidance Implementation teams
Relationship to FSI-CopilotGov Scope boundary with the companion Copilot framework New users, program leads
Solutions Integration Companion solution catalog and automation coverage Implementation teams

Advanced & Scaling

First-run admins: Start with the Core Documents above and the Control Catalog. The four documents below support governance leads and architects who need to map CIO conversations onto zones and controls, assess organizational readiness, or design an Agentic CoE — they are not required for initial control implementation.

Document Purpose Audience
Frontier Transformation Patterns Six deployment patterns mapped to zone defaults and regulatory exposure profiles AI governance leads, architects
Agentic Capability Drivers Five organizational readiness dimensions and maturity targets by pattern AI program sponsors, governance leads
Unified Agent Governance Centralized agent identity, authentication, and authorization (Entra Agent ID, Agent 365, Admin Center) Security architects
Agentic Center of Excellence Four-function CoE blueprint for scaling agent governance without gatekeeper bottlenecks CoE leads, executive sponsors

Framework Principles

1. Risk-Based Governance

Controls scale with risk. Zone 1 (personal productivity) requires minimal oversight while Zone 3 (enterprise/customer-facing) requires comprehensive governance including committee approval, model risk management, and record-class retention as defined in Control 1.7 rather than a blanket retention period.

2. Regulatory Alignment

The framework maps controls to US financial regulations including FINRA 4511/3110, SEC 17a-3/4, SOX 302/404, GLBA 501(b), OCC Bulletin 2026-13 (formerly OCC 2011-12), and Fed SR 26-2 (formerly SR 11-7). Organizations should validate mappings against their specific regulatory obligations.

3. Microsoft Platform Foundation

All controls leverage native Microsoft 365 and Power Platform capabilities. This framework does not require third-party governance tools, though organizations may integrate additional solutions.

4. Separation of Concerns

The framework separates:

  • Framework (this layer) — Stable governance principles updated 1-2x per year
  • Control Catalog — Control objectives and requirements updated quarterly
  • Playbooks — Implementation procedures updated continuously as Microsoft portals change

Quick Navigation

For Executives:

  1. Start with Executive Summary
  2. Review Zones and Tiers for risk classification
  3. Understand Operating Model for accountability

For New Users Comparing Frameworks:

  1. Read Relationship to FSI-CopilotGov to confirm scope
  2. Review Zones and Tiers for the AgentGov operating model
  3. Continue to Executive Summary for governance context

For Compliance Officers:

  1. Review Regulatory Framework
  2. Understand Governance Cadence for examination readiness
  3. Reference Control Catalog for specific requirements

For Implementation Teams:

  1. Run the Governance Readiness Assessment to identify gaps and priorities
  2. Follow Adoption Roadmap
  3. Reference Playbooks for step-by-step procedures
  4. Use Control Catalog for control objectives

Version Information

  • Framework Version: 1.6.2 (May 2026)
  • Last Updated: May 2026
  • Update Frequency: 1-2 times per year (major regulatory or platform changes)


Next → Executive Summary — Board-level overview of AI agent risks, governance model, and investment requirements.


FSI Agent Governance Framework v1.6.2 - May 2026