Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
The Identity Risk Reduction Agent is a green team agent in Project Perception that helps organizations reduce identity risk. The agent assesses how your users are protected. It considers how they authenticate, which Conditional Access policies apply to them, and which identity risk signals Microsoft Security has observed. For each gap, it recommends a specific identity protection and explains why. The agent doesn't directly modify your environment. Administrators review and approve recommended changes through Project Perception, which coordinates the execution of approved actions.
The agent provides the identity expertise, and playbooks give it an objective. Each playbook points the agent at a specific identity risk question, such as moving users to phishing-resistant authentication or closing identity attack paths that were already used against your users. In preview, the agent focuses on user identities in Microsoft Entra ID and on protections enforced through Conditional Access.
Use this article to choose a playbook, review its findings, and approve or reject recommendations as an identity administrator. Before you begin, review the prerequisites for access, licensing, and permissions.
Important
Some information in this article relates to a prereleased product which may be substantially modified before it's commercially released. Microsoft makes no warranties, expressed or implied, with respect to the information provided here. Project Perception is currently in Limited Public Preview. Limited Public Preview Program means Perception is available to a small, invitation-only set of customers for a defined preview window before broader public availability.
Why use the Identity Risk Reduction Agent?
The Identity Risk Reduction Agent addresses the identity side of posture hardening. Identity is one of the most common ways attackers gain access, and most organizations already know which protections reduce that risk. The challenge is applying them to the right users, at the right time, without disrupting the business. That's true whether you're proactively moving users to phishing-resistant authentication or responding to identity attacks seen in your environment.
In both cases, doing the work manually means pulling together how users authenticate, which Conditional Access policies apply to them, and, when relevant, the identity risk Microsoft Entra has detected. Then you have to work out which users need the protection, who already has it, and which policy would close the gap.
The agent does that work for you. It doesn't replace Microsoft Entra ID, Conditional Access, Microsoft Entra ID Protection, or your existing identity tools. Instead, it reasons across their data to find where users lack the protection they need, scope each gap to the users it affects, and turn it into a change you can review and approve. Administrators can review and approve or reject these recommendations through Project Perception before any changes are applied.
The same agent supports proactive hardening, such as rolling out phishing-resistant authentication, and threat-informed hardening, such as closing an attack path that was used against your users. The playbook you run sets the objective.
Use the agent to:
- Find identity exposure. See where users lack the protection they need, whether because of weak authentication or a gap in Conditional Access coverage.
- Target the right users. Get recommendations scoped to the users who need them, and see which users are already covered, so you change only what needs changing.
- Understand the decision. Review the evidence behind each recommendation: the users in scope, their current protection, and the control that addresses the gap.
- Move to action safely. Review and approve or reject recommended changes through Project Perception. The platform coordinates approved actions, and no changes are applied without administrator approval.
- Turn goals and evidence into action. Whether you start from a posture goal or from attack evidence, the agent produces a specific, scoped change that identity administrators can review and act on.
Scenarios
The agent supports both proactive hardening and threat-informed hardening. Each scenario runs as its own playbook.
| If your question is... | Use this playbook | What it does |
|---|---|---|
| "How do we reduce credential phishing risk across our organization?" | Reduce credential phishing risk | Assesses readiness for phishing-resistant authentication, such as passkeys, and recommends a Conditional Access policy that requires it for users who are ready. |
| "Our users were targeted by identity attacks like device code phishing or AiTM. Are they protected now?" | Close identity attack paths used in your environment | Uses identity risk detections in your tenant to find users targeted by device code phishing (DCF) or adversary-in-the-middle (AiTM) attacks, checks their Conditional Access coverage, and recommends policies to close the gaps. No input is required. |
For information about agents and playbooks, see Key concepts in Project Perception and Work with playbooks.
Reduce credential phishing risk is the starting point for ongoing hardening across your organization. Close identity attack paths used in your environment is the better choice when attacks have already been seen in your tenant and you need to check whether the targeted users are protected.
Note
These are distinct flows with different outputs. Reduce credential phishing risk evaluates authentication readiness without a threat filter. Close identity attack paths used in your environment starts from risk detections in your tenant and evaluates only the users associated with them.
Note
During preview, playbook names in Project Perception might differ from those used in this article:
- Close identity attack paths used in your environment might appear as Threat-Informed Posture Hardening.
- Reduce credential phishing risk might appear as Phishing-resistant MFA.
For related posture guidance, see Posture Prioritization Agent.
Before you begin
Before you use the Identity Risk Reduction Agent, confirm product availability, licensing, permissions, and configuration. Ensure you have:
- Access to Project Perception and the permissions required to set up or run the applicable playbook. For more information, see Get started with Project Perception.
- The Security Administrator role to set up the agent and to approve the changes it recommends.
- The required Microsoft 365 E7 licensing.
Each playbook also needs data in your tenant:
- Reduce credential phishing risk: Users assigned to at least one supported built-in Microsoft Entra admin role. If every in-scope admin is already covered by an enabled Conditional Access policy that requires phishing-resistant multifactor authentication (MFA), the playbook finishes without recommending changes.
- Close identity attack paths used in your environment: Microsoft Entra ID Protection risk detections related to device code phishing (DCF) or adversary-in-the-middle (AiTM) attacks. If your tenant has no relevant detections, the playbook completes without a recommendation.
Important
The agent analyzes the identity data available in your tenant when the session runs. A session without recommendations might indicate that no users met the playbook's criteria, that users in scope already have the recommended protection, or that relevant attack detections weren't available. It doesn't mean your tenant is protected against every identity attack.
Tip
Emergency-access (break-glass) accounts are not automatically excluded. Administrators must select the accounts they want to exclude as a playbook input.
Set up the Identity Risk Reduction Agent
Follow the steps in Set up an agent to make the agent available in Perception. During setup, Microsoft Entra creates a dedicated Agent ID for the Identity Risk Reduction Agent, and you review the permissions the agent needs. The agent reads data with the permissions assigned to its own identity, not those of the person who starts a session.
The Identity Risk Reduction Agent runs as part of a playbook. A single agent can be used by multiple playbooks, so after you set up the agent, every playbook that uses it becomes available. Choose the playbook that matches the question you want to answer.
For information about managing agents, see Work with agents.
Start an Identity Risk Reduction Agent session
For ways to start a session, see Start a new session. Select the playbook that matches your current need.
Reduce credential phishing risk
Use this playbook when you want to reduce exposure to credential phishing by requiring phishing-resistant authentication methods, such as passkeys.
- Go to Playbooks.
- Select the Phishing-resistant MFA playbook.
- Select New session.
- Select Start session.
Close identity attack paths used in your environment
Use this playbook when you want to check whether users targeted by supported identity attacks are covered by the right Conditional Access controls. No input is required.
- Go to Playbooks.
- Select the Threat-Informed Posture Hardening playbook.
- Select New session.
- Select Start session.
Monitor session progress
When a session starts, the agent creates a session in Project Perception. To view and manage sessions:
- Go to Agents.
- Under Agents in use, select the Identity Risk Reduction Agent.
- Under Recent Sessions, select the session by timestamp.
The Identity Risk Reduction Agent page includes the following tabs:
| Tab | Description |
|---|---|
| Overview | Shows the agent's status, playbooks, identity, permissions, and recent sessions. |
| Performance | Shows how the agent is used, including session activity, execution status, compute consumption, and approvals. |
Understand the report
When a session completes, the agent provides an assessment summary. The summary is also available as an output file in the session's Outputs section.
The details depend on the playbook you ran, but every report helps you answer the same questions: which users are in scope, whether they're protected, and what the agent recommends.
Start with the summary
Read the summary first. It describes what the agent assessed, how many users are in scope, and how many of them need coverage.
Review the findings
For each recommendation, review:
- Recommended control: The protection that addresses the gap. For example, Require phishing-resistant MFA or Block device code flow.
- Users in scope: The users the playbook's objective applies to.
- Covered users: Users already protected by an existing policy that applies the recommended control. These users don't need a change.
- Users needing coverage: Users who aren't protected by the recommended control. The recommendation targets these users.
- Users that need review: Users whose coverage the agent couldn't confirm, for example because of exclusions, conditional assignments, or conflicting policies. Review these users manually.
- Evidence: The signals and existing policies the agent used to reach its conclusion.
Both Reduce credential phishing risk and Close identity attack paths used in your environment display results in the same format and grouping. Results are grouped by attack pattern: device code phishing and AiTM.
Note
In preview, affected users might be shown by their Microsoft Entra object ID rather than by display name.
Approve recommendations
For the Close identity attack paths used in your environment playbook, the session runs while the agent builds its assessment. When the agent has a recommendation, the session status changes from In Progress to Waiting for input. The agent posts an approval request.
- Review the recommendation and select one of the following:
- Approve to have Perception apply the recommended change.
- Reject to decline the recommendation. Nothing is changed.
- Review the recorded response.
A session can include more than one recommendation. Review and respond to each individual approval request.
For more information about approvals, see Approve, reject, or stop agentic work.
Verify the result in Microsoft Entra
After approving a recommendation, go to Microsoft Entra admin center > Conditional Access > Policies and locate the newly created policy. The policy name should be stated in the assessment output.
The policy is initially created in Report-only mode, so the recommended protection isn't yet enforced. When you're ready to enable protection, you have two options:
- Enable the policy directly in Microsoft Entra: Open the policy, change Enable policy from Report-only to On, and select Save.
- Enable through the playbook: Run the playbook again to receive a recommendation to enable the existing policy. Review and approve the recommendation through the Perception approval workflow.
Policies aren't automatically enabled after 30 days.
When there's no recommendation
If the agent doesn't find users in scope who need coverage, the session completes without an approval request. This happens when no users meet the playbook's criteria, or when all users in scope are already covered.
Note
Summaries and explanations are AI-generated and might be incomplete or incorrect. Changes are made only through the approval step, not by the model on its own. Review each recommendation against your organization's change process before you approve it.
Data coverage and limitations
The following scope and limitations apply during preview:
| Scope | Details |
|---|---|
| User identities only | The agent analyzes user identities. Workload identities and AI agent identities aren't supported. |
| Microsoft Entra data | The agent uses Microsoft Entra data, such as authentication methods, Conditional Access configuration, and ID Protection risk detections. Third-party identity signals aren't supported. |
| Conditional Access protections | The agent's recommended changes are Conditional Access policies. Changes are limited to what the agent recommends and an administrator approves. |
| Approval required | The agent recommends changes but doesn't apply them directly. Administrators review and approve or reject each recommendation through Project Perception, which coordinates the execution of approved actions. |
| Playbook scope | Reduce credential phishing risk applies to users assigned to at least one supported built-in Microsoft Entra admin role. If every in-scope admin is already covered by an enabled Conditional Access policy that requires phishing-resistant MFA, the playbook finishes without recommending changes. Close identity attack paths used in your environment supports device code phishing and AiTM phishing. |
| Point-in-time analysis | Results reflect the data available when the session ran. Run the playbook again to see updated coverage. |
| Policy enforcement | Conditional Access policies created through the Close identity attack paths used in your environment playbook are initially set to Report-only mode. Policies aren't automatically enabled. Administrators can enable them directly in Microsoft Entra or run the playbook again to review and approve a recommendation to enable an existing policy. |