Frequently Asked Questions (FAQ)
Common questions and answers about the FSI Agent Governance Framework.
Getting Started
Q: Where do I start?
A: Read in this order:
- README.md (5 min overview)
- Zones Guide (understand your zone)
- Quick Start (30 min hands-on)
- Regulatory Mappings (if regulated)
Q: What if I don't know which zone my agent should be in?
A: Use the Zone Decision Matrix:
- Zone 1: Personal only, M365 data only
- Zone 2: Team/dept, internal data only
- Zone 3: Org-wide, customer data, regulated data
Ask your manager or compliance officer if unsure.
Q: How long does implementation take?
A: Depends on current state:
- Zone 1: 1-2 days (minimal)
- Zone 2: 1-2 weeks (moderate)
- Zone 3: 3-6 weeks (comprehensive)
Full framework: 8-week phased approach (see the Implementation Checklist)
Framework Questions
Q: Why are there 48 controls?
A: The framework covers:
- Pillar 1: 19 security controls
- Pillar 2: 15 management/lifecycle controls
- Pillar 3: 9 reporting/monitoring controls
- Pillar 4: 5 SharePoint-specific controls
Total = 48 controls covering all governance areas.
Q: Do I need to implement all 48 controls?
A: No. Implement based on:
- Your zone: Zone 1 needs fewer, Zone 3 needs all
- Your regulations: Implement applicable controls
- Your risk tolerance: Higher risk = more controls
Start with baseline for your zone, add recommended/regulated as you mature.
Q: What's a "governance level"?
A: Each control has 3 levels:
- Baseline (Level 1): Minimum implementation
- Recommended (Level 2-3): Best practices
- Regulated/High-Risk (Level 4): Comprehensive
Implement baseline for Zone 1, baseline+recommended for Zone 2, all levels for Zone 3.
Q: Can I modify the framework?
A: Yes! The framework is a starting point. You can:
- Add controls specific to your organization
- Combine controls if makes sense
- Adjust governance levels based on risk
- Customize for your regulations
Maintain version control and document changes.
Environment Governance
Q: How do I prevent users from creating agents in the default environment?
A: Enable environment routing in Power Platform Admin Center:
- Navigate to PPAC → Manage → Environment groups
- Enable Default environment routing
- Configure routing rules to direct makers to appropriate environments
- Optionally enable Developer environment auto-provisioning
This prevents shadow AI by automatically routing makers to governed environments. See Control 2.15: Environment Routing.
Q: Where can I see all agents across my tenant?
A: Use the PPAC Inventory experience (Preview):
- Power Platform Admin Center → Resources → Agents
- View tenant-wide agent list with metadata
- Filter by environment, owner, or status
- Note: Data refreshes every 24 hours; 500 agent display limit
Also check M365 Admin Center → Settings → Integrated Apps for published agents.
See Control 3.1: Agent Inventory.
Q: How do I promote an agent from Zone 2 to Zone 3?
A: Follow the zone promotion process:
- Complete all Zone 3 governance requirements
- Submit formal promotion request to Governance Committee
- Undergo security and compliance review
- Configure production environment with Managed Environment enabled
- Use ALM pipelines to deploy from test to production
- Document promotion and retain evidence
See Lifecycle Governance Guide and Control 2.3: Change Management.
Q: What are Environment Groups and how do they help governance?
A: Environment Groups allow you to:
- Group environments by zone (Zone 1, Zone 2, Zone 3)
- Apply consistent governance rules across all environments in a group
- Enforce connector policies, sharing limits, and AI model restrictions
- Prevent configuration drift with centralized rule management
Navigate to PPAC → Manage → Environment groups to configure. See Control 2.2: Environment Groups.
Q: What is the Copilot Hub?
A: The Copilot Hub (in Power Platform Admin Center) provides:
- Centralized dashboard for agent governance
- Usage and adoption metrics
- ROI tracking and business value insights
- Capacity/consumption monitoring
- Quick access to governance controls
Access via PPAC → Copilot. See Control 3.8: Copilot Hub.
Q: How do I monitor AI data security risks?
A: Use Microsoft Purview DSPM for AI:
- Navigate to purview.microsoft.com → DSPM for AI
- Review recommendations for AI security
- Enable activity monitoring for AI interactions
- Configure DLP policies targeting AI applications
- Run oversharing assessments for agent knowledge sources
Q: What are the 48 controls?
A: The framework includes 48 controls across four pillars:
- Pillar 1 - Security: 19 controls (1.1-1.19) covering DLP, encryption, audit logging, eDiscovery
- Pillar 2 - Management: 15 controls (2.1-2.15) covering lifecycle, change control, environment routing
- Pillar 3 - Reporting: 9 controls (3.1-3.9) covering inventory, monitoring, incidents, Sentinel
- Pillar 4 - SharePoint: 5 controls (4.1-4.5) covering SharePoint-specific governance
See Control Index for the complete list.
Q: How do I monitor my Power Platform security posture?
A: Use the PPAC Security Posture Assessment:
- Navigate to Power Platform Admin Center → Security → Overview
- Review your security score (Low/Medium/High)
- View security recommendations
- Click recommendations to see remediation steps
- Track improvements over time
This provides a centralized view of tenant security configuration. See Control 3.7: PPAC Security Posture Assessment.
Q: How do I integrate with Microsoft Sentinel for agent monitoring?
A: For Zone 3 agents requiring SOC integration:
- Configure Microsoft Sentinel workspace in Azure
- Enable Power Platform data connector in Sentinel
- Create analytics rules for agent-related security events:
- Unusual agent data access patterns
- Connector policy violations
- Environment configuration changes
- Configure incident response playbooks
- Integrate with your SOC procedures
This enables real-time threat detection and automated response for production agents. See Control 3.9: Microsoft Sentinel Integration.
Zones & Classification
Q: Can an agent be in multiple zones?
A: No. Each agent is in one zone. It may progress through zones:
- Zone 1 (personal) → Zone 2 (team) → Zone 3 (production)
Q: What if an agent starts in Zone 1 but needs Zone 2 features?
A: Promote it through formal process:
- Request promotion to next zone
- Get appropriate approvals
- Implement required controls for new zone
- Move to new environment
- Document promotion
Q: Can I move an agent from Zone 3 to Zone 2?
A: Yes, through demotion if:
- Agent is no longer production-critical
- Compliance requirements reduced
- Risk profile decreased
Requires governance committee approval and documentation.
Q: What defines "customer-facing"?
A: If the agent is used by:
- Customers
- Clients
- External parties
- Regulatory subjects
→ Must be Zone 3
Governance & Approvals
Q: Who approves each zone?
A: - Zone 1: Self-service (no approval) - Zone 2: Manager/Department Head - Zone 3: Governance Committee (Compliance Officer, CISO, General Counsel, CRO)
Q: How long does approval take?
A: - Zone 1: Immediate (self) - Zone 2: 3-5 business days (manager review) - Zone 3: 10-14 business days (governance committee)
Q: What if my approval is denied?
A: Get feedback and revise:
- Address concerns raised
- Provide additional information
- Re-submit
- Escalate if needed per the RACI Matrix
Q: Who is the Governance Committee?
A: Typically:
- AI Governance Lead (Chair)
- Compliance Officer
- CISO
- General Counsel
- Chief Risk Officer (if applicable)
- Business owner (agent requester)
See the RACI Matrix for detailed roles.
Regulations
Q: Which regulations apply to my organization?
A: - FINRA? If you're a securities broker/dealer - SEC? If you're a registered investment adviser or public company - SOX? If you're a public company - GLBA? If you're a financial institution - OCC? If you're a national bank - Federal Reserve? If you're a bank holding company
Check with your Compliance Officer.
Q: What if multiple regulations apply?
A: Implement controls that satisfy all:
- Check Regulatory Mappings for each regulation
- Take strictest requirement
- Document compliance with each
- Implement control at highest level needed
Example: If both FINRA (1yr) and SEC (6yr) apply, implement 6-year retention.
Q: Are there controls I don't need to implement?
A: Possibly. Review Regulatory Mappings:
- Find your regulations
- See applicable controls
- Implement those controls
- Others are optional but recommended
Q: We're not regulated. Do we still need governance?
A: Yes, best practices recommend:
- Basic security (MFA, DLP, audit)
- Change management
- Testing before production
- Incident response
Even without regulations, governance protects:
- Data
- Operations
- Reputation
- Business continuity
Implementation
Q: Should we do all phases at once?
A: No. The 8-week phased approach recommended:
- Phase 1 (Weeks 1-2): Assessment
- Phase 2 (Weeks 3-4): Security baseline
- Phase 3 (Weeks 5-6): Advanced governance
- Phase 4 (Weeks 7-8): Finalization
See the Implementation Checklist for details.
Q: Can we start before full governance is ready?
A: Yes, by zone:
- Zone 1: Deploy immediately (no approval needed)
- Zone 2: Can deploy after basic approval workflow
- Zone 3: Must have full governance before production
Recommend completing Phase 1-2 before Zone 3 deployment.
Q: What if we can't implement all controls by deadline?
A: Prioritize:
- Security controls (Pillar 1) - highest priority
- Regulatory requirements - next priority
- Management controls (Pillar 2)
- Reporting controls (Pillar 3)
Implement basic version for urgent items, mature over time.
Q: Who leads implementation?
A: Typically:
- Project Lead: AI Governance Lead
- Executive Sponsor: Compliance Officer or CISO
- Implementation Team:
- Power Platform Admin (technical)
- Compliance Officer (regulatory)
- CISO or Security Admin (security)
- Internal Audit (independent testing)
See the RACI Matrix for detailed roles.
Operations & Monitoring
Q: How often should we review controls?
A: - Zone 1: Annual (if tracked) - Zone 2: Quarterly - Zone 3: Monthly + annual comprehensive
Q: What's the difference between monitoring and assessment?
A: - Monitoring: Continuous (daily/weekly) tracking of activity - Assessment: Periodic (quarterly/annual) evaluation of control effectiveness
Both are needed.
Q: How do we measure compliance?
A: Use maturity scorecard:
- Level 0 (0%): Not implemented
- Level 1 (25%): Baseline implemented
- Level 2 (50%): Developing toward recommended
- Level 3 (75%): Recommended implemented
- Level 4 (100%): Regulated/high-risk implemented
Track progress over time.
Q: What if we find a compliance gap?
A: Follow incident management:
- Document the gap
- Assess severity and impact
- Create remediation plan
- Assign owner and deadline
- Verify remediation
- Document closure
See Control 3.4: Incident Reporting for details.
Technology & Platforms
Q: What platforms does this framework support?
A: Microsoft 365 agents and related in-suite Copilot experiences, including:
- Copilot Studio agents
- Agent Builder agents
- SharePoint agents
- Teams agents
- Microsoft 365 Copilot
Q: Do we need specific licenses?
A: Depends on controls:
- Basic governance: Standard M365 licenses
- Advanced governance: Premium licenses recommended for:
- Purview Audit Premium (longer retention)
- Managed Environments (governance enforcement)
- DSPM for AI (data governance)
Check with your Microsoft account team.
Q: What about Agent 365 and Entra Agent ID?
A: Some controls reference Microsoft features that are currently in preview:
| Feature | Status (Dec 2025) | Access |
|---|---|---|
| Agent 365 | Frontier Preview | Requires Frontier program enrollment |
| Entra Agent ID | Public Preview | Available in Entra Admin Center |
| Advanced Connector Policies (ACP) | Preview | Available in PPAC |
| Environment Groups | Preview | Available in PPAC |
To access preview features:
- Frontier program: Sign up at the Microsoft 365 Admin Center → Settings → Org settings → Frontier
- Agent ID: Navigate to Entra Admin Center → Enterprise applications → Filter by "Agent ID (Preview)"
Controls that reference preview features include appropriate disclaimers. Check Microsoft Learn for current availability.
Q: Can we use other governance platforms?
A: Yes. Framework is platform-agnostic. You can:
- Use ServiceNow for change management
- Use Jira for incident tracking
- Use custom compliance tools
- Mix and match tools
Important: Ensure integration and audit trail.
Audit & Compliance
Q: How long do we keep records?
A: Depends on regulation:
- FINRA 4511: 6 years + 1 year accessible
- SEC 17a-3/4: 6 years + 3 years accessible
- SOX 404: 7 years minimum
- GLBA: 5-7 years
- OCC/SR 11-7: Per model (typically 3+ years)
See Regulatory Mappings for your regulations.
Q: What evidence do auditors want to see?
A: For each control:
- Policy documentation (what should happen)
- Configuration proof (technical setup screenshots)
- Activity logs (evidence it's working)
- Test results (verification that it works)
- Remediation records (how issues were fixed)
Q: How do we prepare for an audit?
A: Complete preparation process:
- Inventory all agents (Control 3.1)
- Document policies and procedures (Control 2.13)
- Compile audit evidence (activity logs, approvals, tests)
- Perform control testing (internal audit)
- Remediate any gaps (document remediation)
- Create compliance summary (executive overview)
Q: What if auditors find a violation?
A: Remediation process:
- Root cause analysis: Why did it happen?
- Corrective action: Fix the issue
- Preventive action: Stop recurrence
- Documentation: Record everything
- Evidence: Verify remediation worked
- Reporting: Report to audit committee
Exceptions & Risk Acceptance
Q: What if we cannot implement a required control?
A: Document an exception with risk acceptance:
- Document the gap: Identify the control and what cannot be implemented
- Explain the reason: Technical limitation, business constraint, or cost prohibition
- Assess the risk: What is the potential impact of non-implementation?
- Identify compensating controls: What alternative measures reduce the risk?
- Obtain approval: Get sign-off from appropriate authority (see below)
- Set review date: Schedule periodic re-evaluation
Q: Who must approve a control exception?
A: Approval authority depends on the zone and control criticality:
| Zone | Non-Critical Controls | Critical Controls |
|---|---|---|
| Zone 1 | Manager | AI Governance Lead |
| Zone 2 | AI Governance Lead | Governance Committee |
| Zone 3 | Governance Committee | Governance Committee + Executive Sponsor |
Critical controls include: DLP (1.5), Audit Logging (1.7), MFA (1.11), Access Control (2.8).
Q: What are compensating controls?
A: Alternative measures that reduce risk when the primary control cannot be implemented:
| Primary Control Gap | Possible Compensating Controls |
|---|---|
| Automated DLP not available | Manual review process, restricted data access |
| Environment routing not enabled | Weekly audit of default environment, cleanup procedures |
| Sentinel integration not licensed | Enhanced manual log review, third-party SIEM |
| Managed Environments not licensed | More frequent manual audits, stricter approval process |
Document compensating controls with evidence of their effectiveness.
Q: How long can an exception remain open?
A: Exception durations:
- Temporary exceptions: Maximum 90 days, then re-evaluate
- Long-term exceptions: Maximum 12 months, requires annual renewal
- Permanent exceptions: Rare, requires Governance Committee approval and annual attestation
All exceptions should include a remediation target date when feasible.
Q: How do we document exceptions for auditors?
A: Maintain an Exception Register with:
- Control ID and name
- Exception description (what is not implemented)
- Business justification
- Risk assessment (likelihood × impact)
- Compensating controls (with evidence)
- Approval record (who approved, when)
- Review schedule (next review date)
- Remediation plan (if applicable)
Auditors expect documented risk acceptance decisions, not undocumented gaps.
Troubleshooting
Q: Our DLP policy isn't working.
A: Check:
- Is the policy enabled?
- Is it scoped to the right locations?
- Is the rule correctly configured?
- Are there any exceptions?
- Have you tested with sample data?
See Control 1.5 for troubleshooting steps.
Q: We're seeing too many false positives on DLP.
A: Tune your DLP:
- Review rule sensitivity
- Add exceptions for legitimate uses
- Adjust thresholds
- Test with realistic data
- Document exceptions
Q: Audit logs seem incomplete.
A: Verify:
- Is audit logging enabled?
- Is retention policy set?
- Are all locations being logged?
- Check for export/archival jobs
- Verify access permissions to audit logs
Q: We missed a control deadline.
A: Reassess:
- What's the current state?
- What's required by your zone/regulation?
- Can you implement a baseline version quickly?
- Create revised timeline
- Document re-plan and approval
Best Practices
Q: What's the #1 governance mistake?
A: Not starting with clear policies. - Define governance objectives - Document approval procedures - Communicate expectations - Technology is secondary to process
Q: How do we get buy-in?
A: Key stakeholders:
- Business: Show governance enables innovation (Zone 1 is fast)
- Compliance: Show you're meeting requirements
- IT: Show you're supporting their platform
- Executives: Show you're reducing risk
Q: How do we scale governance?
A: As agents grow:
- Automate where possible (DLP, audit)
- Delegate approvals (managers for Zone 2)
- Create templates (agent request forms)
- Use dashboards (real-time monitoring)
- Document procedures (make it repeatable)
Getting Help
Q: Where do I find details on a specific control?
A: Reference the control file:
- Example: "1.5-data-loss-prevention-dlp-and-sensitivity-labels.md"
- Each control has implementation guidance and verification steps
Q: Who can I contact with questions?
A: - Governance: Contact AI Governance Lead - Regulatory: Contact Compliance Officer - Technical: Contact Power Platform Admin - Security: Contact CISO or Security Admin
Q: Is there a glossary?
A: Yes! See Glossary.md for all terms and definitions.
Q: Where can I get training?
A: See Control 2.14: Training & Awareness Program - Role-specific training available - Annual refresher required - Compliance certification recommended
FSI Agent Governance Framework Beta - December 2025