My first look at Intune Agents (part1)

Unless you've been sleeping for the past year you'll have heard about Microsoft's Copilot offering. There are different flavours: M365 Copilot, Security Copilot, GitHub Copilot etc. I'm an Intune guy, so I'm mostly interested in Security Copilot. In this blog post I'll discuss how to get started and my first look at the Intune Agents, which are in Public Preview and use Security Copilot under the hood.

First the prerequisites for Security Copilot, there are a few.

• You need an Azure subscription in order to provision Security Compute Units (SCUs); more about that below.
• You need an account which has been assigned the correct role to configure Copilot capacity (SCUs); also more about that below.

Security Compute Units

Security Compute Units (SCUs) are the compute capacity required to run Security Copilot workloads. 

At Ignite 2025 Microsoft announced that Security Copilot will be available to all Microsoft 365 E5 customers. The rollout starts November 18, 2025, for existing Security Copilot customers, and will continue in the upcoming months for all Microsoft 365 E5 customers. What does that mean? Customers with Microsoft 365 E5 will have 400 Security Compute Units (SCU) each month for every 1,000 paid user license, up to 10,000 SCUs each month. So, an organization with 4,000 user licenses gets 1,600 SCUs/month. This is great news and has made Security Copilot more affordable for organizations. It's important to note that the cost for M365 E5 licenses has increased at this time, but Microsoft have also added the Intune Suite features to the E5 subscription.

How will you know how many SCUs are being consumed? Security Copilot provides a usage monitoring dashboard for Copilot owners, allowing them to track usage over time. We'll have a look at that later. 

Microsoft Entra and Microsoft Purview roles

The following Microsoft Entra and Microsoft Purview roles automatically inherit Copilot owner access:

Microsoft Entra roles:

• Billing Administrator
• Entra Compliance Administrator
• Global Administrator
• Intune Administrator
• Security Administrator

Microsoft Purview roles:

• Purview Compliance Administrator
• Purview Data Governance Administrator
• Purview Organization Management

Once Security Copilot is rolled out to your organization and available via your M365 E5 licensing, then SCUs will automatically be available. If you don't qualify through your licensing, then you will have to provision SCUs yourself. The Microsoft documentation shows you how to get started with that.

This is done by navigating to https://securitycopilot.microsoft.com where you are guided through the steps.

Tip: SCUs are provisioned on an hourly basis. To maximize usage, make SCU provisioning changes at the beginning of the hour.

Intune Agents

Once SCUs have been provisioned, then Security Copilot is available for your organization. This lights up Agents in the Intune portal.

For now, three Agents are available in the Public Preview.

• Change Review Agent: uses Microsoft Security Copilot's generative AI to evaluate Multi Admin Approval requests for PowerShell scripts on Windows devices. It provides risk-based recommendations and contextual insights to help administrators understand script behaviour and associated risks. I'll be concentrating on this agent for this blog post.
• Device Offloading Agent: identifies stale or misaligned devices across Intune and Entra ID, providing actionable insights and offboards devices subject to admin approval.
• Policy Configuration Agent:  helps IT admins to translate complex requirements and industry standard documents into actionable Intune settings, and allows administrators to quickly generate Intune settings catalog policies. 

Ok, let's look more closely at the Change Review Agent.

Click on View details.

You are prompted to Set up agent.

After the agent is set up, click Run to start a job. This should examine my PowerShell scripts (that are subject to Multi-Admin approval) and to identify and risks associated with the script.

My first run didn’t identify any suggestions, which surprised me. I had previously uploaded a script which was subject to multi-admin approval. The script had been approved by a second administrator. This script is very destructive as it resets any Windows device back to factory settings. I ran the agent and expected to see some suggestions about the script but received nothing. 

Then I figured out that the agent was only targeted to "pending" requests, not requests that have actually been "approved". I uploaded the script again without approving and ran the agent again. 

This time I received a suggestion called Reject Device reset. I clicked on the suggestion and was very impressed with the output. The agent figured out exactly what this script would do and told me that I should reject the approval. Then I knew that I had to be very careful with this script that someone else created.

 

Suggested action: Reject

The script is designed to perform a remote wipe operation on devices using WMI, which is a highly destructive action. While there are no explicit signs of malicious code or prior security rejections, the script's purpose and actions involve significant risk due to the potential for data loss and system disruption. The business justification is minimal, and the script's risk level is high based on its destructive capability. All validation points except Script Reputation have sufficient data, but the lack of reputation data does not mitigate the high inherent risk of the script's function.

The script interacts with the Windows Management Instrumentation (WMI) to invoke a remote wipe method on a device. It connects to the `root\cimv2\mdm\dmmap` namespace and targets the `MDM_RemoteWipe` class. The method `doWipeMethod` is executed with specific parameters, potentially to perform a remote wipe operation. The script includes error handling to capture and display any exceptions that occur during the process.

Factors:

• Script Purpose: The script's primary function is to execute a remote wipe, which is a destructive operation. Although the actions are well-documented, the risk of unintended or malicious use is high. The metrics require no risky constructs and a well-defined scope, but the destructive nature of the script outweighs these controls for a Create operation.
• Approval/Rejection History: There is no history of prior rejections or security risk decisions for this script, so this point is satisfied.
• Alert History (Script): No high-severity alerts or active incidents are associated with the script, meeting the criteria for this point.
• Business Justification: The justification is minimal and lacks detail on scope, controls, and privacy. For a high-risk operation like remote wipe, a more comprehensive justification is required.
• Requestor Risk Indicators: The requestor is not marked as deleted or risky, and there are no unresolved risk indicators, so this point is satisfied.

Very accurate, really clever stuff.

Finally browse to Security Copilot Usage Monitoring to see how many SCUs are being used.

I hope you found this useful. Until next time.......

See my follow up post "My first look at Intune Agents (part2)"