6.1. Automate Management Tasks
💡 First Principle: Anything you do in the admin center changes an Intune object, so code that reads and writes the same object can do it too. Automation adds no new powers: it removes the clicks and the human timing, while licensing, permissions and approvals still apply. So the real question is who the code acts as, and what that identity may do.
Why care? The exam tests whether you can match each repeatable job — a Monday compliance export, a registry value no setting covers, a backlog of Defender vulnerabilities — to the right tool and permission.
Think of a ladder of judgment, from scripts where you write every decision to an AI agent that ranks options but doesn't act:
| Tool | Where the logic runs | Who decides | Covered in |
|---|---|---|---|
| Microsoft Graph and Graph PowerShell | Admin workstation, automation account or registered app | You, in the script | 6.1.1 |
| Custom compliance discovery script | On the device, every eight hours | Intune, comparing output to your JSON rules | 6.1.2 |
| Remediations | On the device, on your schedule | Your detection script's exit code | 6.2.3 |
| Vulnerability Remediation Agent | In Security Copilot, against Defender data | You, after reading its suggestions | 6.1.3 |
⚠️ Common Misconception: The word "agent" makes many people believe Security Copilot agents in Intune change devices on their own. They don't: an admin starts the Vulnerability Remediation Agent manually, and it returns prioritized suggestions only. The admin makes the fix with normal Intune tools, such as a quality update policy or an app update, then marks the suggestion as applied; the agent's Microsoft Entra agentic identity holds only the read permissions you delegate.