Control 3.9: Microsoft Sentinel Integration - Portal Walkthrough
This playbook provides step-by-step portal configuration guidance for Control 3.9.
Prerequisites
- Microsoft Sentinel workspace deployed
- Microsoft Sentinel Contributor or higher role
- Log Analytics Workspace Contributor
- Microsoft 365 E5 or Microsoft Sentinel standalone license
Step 1: Deploy Microsoft Sentinel Workspace
Portal Path: Azure Portal > Microsoft Sentinel
- Navigate to Azure Portal
- Search for "Microsoft Sentinel"
- Click + Create
- Select or create a Log Analytics workspace
- Click Add to enable Sentinel
Step 2: Connect Data Sources
Portal Path: Microsoft Sentinel > Configuration > Data connectors
Connect the following data connectors for agent monitoring:
| Connector | Data Type | Setup |
|---|---|---|
| Microsoft 365 Defender | Security alerts, incidents | One-click setup |
| Microsoft Entra ID | Sign-in and audit logs | Enable diagnostic settings |
| Office 365 | Exchange, SharePoint, Teams activity | Enable |
| Microsoft Defender for Cloud Apps | Cloud app activity | API connection |
| Azure Activity | Azure resource operations | Enable |
Enable Microsoft 365 Defender Connector:
- Navigate to Data connectors
- Search for "Microsoft 365 Defender"
- Click Open connector page
- Click Connect incidents & alerts
- Enable all event types
Step 3: Configure Analytics Rules for Agents
Portal Path: Microsoft Sentinel > Configuration > Analytics
Create custom analytics rules for AI agent monitoring:
Rule 1: Unusual Agent Data Access
| Setting | Value |
|---|---|
| Name | Unusual Agent Data Access Pattern |
| Tactics | Discovery, Collection |
| Severity | Medium |
| Query | See KQL below |
| Run frequency | 5 minutes |
| Lookup data | 1 hour |
Rule 2: Agent DLP Violation
| Setting | Value |
|---|---|
| Name | AI Agent DLP Policy Violation |
| Tactics | Exfiltration |
| Severity | High |
| Query | See KQL below |
| Run frequency | 5 minutes |
Rule 3: After-Hours Agent Activity
| Setting | Value |
|---|---|
| Name | Agent Activity Outside Business Hours |
| Tactics | Initial Access |
| Severity | Low |
| Query | See KQL below |
Step 4: Create Agent Activity Workbook
Portal Path: Microsoft Sentinel > Threat management > Workbooks
Create a workbook with these sections:
| Section | Visualizations |
|---|---|
| Overview | Total agents, active today, alerts by severity |
| Activity Timeline | Agent interactions over time |
| Top Agents | Most active agents ranked |
| Data Access | Documents/sources accessed by agents |
| Anomalies | Flagged unusual activity |
| DLP Events | Policy violations by agent |
Step 5: Configure Automated Responses
Portal Path: Microsoft Sentinel > Configuration > Automation
Create automation rules:
| Rule | Trigger | Actions |
|---|---|---|
| Suspend High-Risk Agent | High severity alert on agent | Disable agent, notify security |
| Escalate DLP Violation | DLP alert for Zone 3 agent | Create incident, email compliance |
| Log Agent Incident | Any agent-related alert | Log to incident tracker |
Step 6: Configure Copilot Studio Application Insights (Custom Telemetry Path)
Portal Path: Copilot Studio > Agent > Settings > Advanced
If using the Application Insights custom telemetry path for conversation-level monitoring (see Control 3.9 "Custom Integration for Comprehensive Telemetry"), configure the following prerequisites:
6a: Enable Application Insights in Agent Settings
- Open Copilot Studio
- Select the target agent
- Navigate to Settings (gear icon) > Advanced
- Click Application Insights
- Enter the Application Insights Connection String
- Enable "Log activities" — this enables basic telemetry flow
- Enable "Log sensitive activity properties" — this includes actual conversation text, user IDs, and node details in the telemetry
Without Step 7
If only "Log activities" is enabled, Application Insights will receive events but conversation content fields will be empty or sanitized. This does not satisfy FINRA 4511 recordkeeping requirements.
6b: Verify Environment-Level Transcript Setting
- Open Power Platform Admin Center
- Navigate to Environments > select the target environment
- Click Settings > Product > Features
- Verify "Allow conversation transcripts" is On
Scope
This setting primarily controls Dataverse transcript storage, but when disabled it can also block downstream telemetry flow to Application Insights.
6c: Verify Data Flow
After enabling both settings, verify conversation data appears in Application Insights:
- Open Azure Portal > Application Insights > select the linked resource
- Navigate to Logs
-
Run a test query to confirm conversation content is present:
customEvents | where timestamp > ago(1h) | where name in ("BotMessage", "UserMessage") | extend text = tostring(customDimensions.text) | project timestamp, name, text | take 10 -
Confirm the
textcolumn contains actual conversation content (not empty values)
Step 7: Set Up Hunting Queries
Portal Path: Microsoft Sentinel > Threat management > Hunting
Create saved hunting queries:
| Query Name | Purpose |
|---|---|
| Agent Data Exfiltration Patterns | Look for bulk data access |
| Dormant Agent Reactivation | Detect unused agents starting |
| Cross-Zone Data Movement | Agent accessing higher-zone data |
| Failed Authentication Patterns | Repeated agent auth failures |
Step 8: Configure Incident Management
Portal Path: Microsoft Sentinel > Threat management > Incidents
Configure incident workflow:
- Set auto-assignment rules for agent incidents
- Configure severity escalation timeline
- Enable entity mapping for agents
- Set up integration with incident tracking (Control 3.4)
Validation
After completing the configuration, verify:
- Microsoft Sentinel workspace deployed and accessible
- Required data connectors enabled and ingesting data
- Analytics rules created and enabled for agent monitoring
- Agent activity workbook displaying data correctly
- Automation rules triggering on test alerts
- Application Insights receiving conversation content (if custom telemetry path configured)
- Hunting queries saved and returning expected results
- Incident assignment workflow functioning
Expected Result: Agent security events flow to Sentinel, analytics rules generate alerts for suspicious activity, and automated responses execute as configured.
Back to Control 3.9 | PowerShell Setup | Verification Testing | Troubleshooting
Updated: January 2026 | Version: v1.2