← BACK TO RUNBOOKS

Runbook

Deploying Microsoft Entra Conditional Access — From Planning to Enforcement

Design a pilot policy, evaluate real sign-ins in report-only mode and decide when it is safe to enforce.

Difficulty
Intermediate
Estimated time
90 min
Published
Last reviewed
Author
Ryan Shenton
Version
0.1

Overview

Conditional Access turns signals about a sign-in into an access decision. This walkthrough follows a single example policy for a small group of test users accessing Exchange Online. It starts with an explicit scope and a recovery plan, then uses observed sign-ins to decide whether to enforce the policy.

By the end of the exercise, you should have a policy record, a test matrix with expected and observed outcomes, and a documented go or no-go decision. The estimate covers hands-on work. Set a separate observation window long enough to include the pilot’s ordinary access patterns and support hours.

The tenant, group and policy labels are illustrative. Work through the exercise in an approved lab. A production deployment needs a separate review of licensing, roles, existing policies, exclusions and rollback ownership.

Prerequisites

Roles and permissions
  • Use a designated Conditional Access Administrator to create or change the pilot policy. Activate the role through Privileged Identity Management if the tenant uses it.
  • Arrange a log reviewer with access to Microsoft Entra sign-in reports. Confirm that the reviewer can see the relevant Conditional Access policy details before the pilot starts.
Licensing
  • Confirm Microsoft Entra ID P1 or an eligible suite for the users in scope. The sample policy uses a standard MFA grant control and has no risk conditions. Review P2 requirements separately if you introduce Identity Protection risk conditions.
Tools and modules
  • Microsoft Entra admin centre, access to a test mailbox, and separate browser profiles for pilot and control sign-ins.
  • Optional inventory example: Microsoft Graph PowerShell with the Authentication and Identity.SignIns modules, delegated Policy.Read.All consent and an authorised reader in the target tenant.
Environment
  • Use a dedicated test tenant with a licensed pilot group and a test mailbox. Have a comparable account outside the group for the control test.
  • Inventory existing Conditional Access policies and check whether security defaults are enabled. Use a lab where the baseline already supports this exercise. A migration from security defaults requires its own change plan.

Preparation checklist

Procedure

0 of 5 steps complete · 0%

  1. Rehearse emergency access

    Confirm that the tenant has at least two cloud-only emergency access accounts. Follow the organisation’s emergency-access procedures for secure credential storage, monitoring and independent phishing-resistant authentication. Test access using an authorised operator and record the test without recording credentials.

    Check that the pilot scope omits emergency identities. Before any enforcement, verify exclusions for policies that could restrict their access and check that the recovery route has no dependency on the control under test.

    Microsoft Learn · Manage emergency access accounts

  2. Record the policy and pilot scope

    Record one control in the change log. For this example the target is interactive user access to Exchange Online. Choose the correct resource from the target tenant rather than assuming the display name alone identifies it.

    • Example name — LAB | Exchange Online | Require MFA
    • Include — a dedicated lab pilot group with a few registered test users.
    • Control group — a licensed test user outside the pilot; keep administrative and emergency identities outside the pilot.
    • Target resource — Exchange Online in the lab tenant.
    • Grant — require multifactor authentication. Initial policy state — Report-only.
    Optional · read-only policy inventory PowerShell
    # Lab example. Use the directory ID of your authorised test tenant.
    $tenantId = '<lab-tenant-id>'
    Connect-MgGraph -TenantId $tenantId -Scopes 'Policy.Read.All' -ContextScope Process
    try {
        Get-MgContext | Select-Object TenantId, Account
        Get-MgIdentityConditionalAccessPolicy -All |
            Select-Object DisplayName, Id, State | Sort-Object DisplayName
    } finally {
        Disconnect-MgGraph | Out-Null
    }
  3. Create the policy in report-only mode

    In the Microsoft Entra admin centre, open Conditional Access policies and create a policy. Enter the pilot group, selected resource and grant control from the change record. Confirm the state reads Report-only before saving. Reopen the saved policy and record its identifier and assignments.

    Microsoft Learn · Understand report-only results

  4. Test sign-ins and review the policy result

    Use What If to check the pilot and control personas with the relevant client and target app. Then perform actual interactive sign-ins from fresh browser sessions. Open the new entries in Microsoft Entra sign-in logs and find the named policy under the report-only results. Compare the policy result with the overall sign-in result.

    • Pilot with MFA already satisfied — expect the policy to apply and the grant requirement to be satisfied.
    • Pilot without an MFA claim, where the lab allows this test — look for an evaluation that indicates more user action would be required. Existing policies or a previously satisfied claim can change the observed outcome.
    • Control account outside the group — expect this pilot policy to be outside scope. Existing tenant policies can still affect the sign-in.
    • Capture UTC time, target resource, account persona, request or correlation ID, expected result and observed result in the authorised test record.

    Microsoft Learn · Conditional Access What If tool

  5. Make the enforcement decision

    At the end of the observation window, review every unexpected outcome. Approve enforcement only when scope and licensing are confirmed, exclusions and recovery have been verified, the test matrix is complete, and the rollback operator is available. Keep the policy in report-only mode while any condition remains open.

    In the authorised lab window, reopen the existing policy, reconfirm the assignments and switch its state to On for the same pilot group. Record the operator and time. Keep any expansion as a separately reviewed change.

Validation

Verification and expected outcomes

  • Confirm the saved policy ID, state, target resource, group membership and grant control match the change record.
  • Use fresh interactive sign-ins to verify that the pilot can access the resource when MFA is satisfied and that the new policy is outside scope for the control identity.
  • Validate the approved recovery path separately and verify the procedure’s logging or alerting for emergency account use.

Post-deployment checks

Review fresh sign-in logs during the agreed observation window. Compare failures and support reports with the pilot baseline. Record unexpected client behaviour and check that affected users can still complete their ordinary work before considering a wider rollout.

Rollback

  • Identify the pilot policy by its recorded ID and have the designated operator open it through an authorised admin session.
  • Return the changed policy to its recorded pre-enforcement state. For this exercise that is Report-only. Use Off when the approved recovery plan calls for stopping evaluation entirely.
  • Confirm the saved state. Retest using fresh sign-ins for affected personas and review new logs. Preserve the policy and change evidence for investigation.
  • Record the rollback time, impact and outcome, then close or escalate the change according to the test results.

Troubleshooting

Known limitations

Report-only evaluates the policy without enforcing its grant. What If is a simulation, so validate access with real pilot sign-ins, including the clients and resources your users actually use. A cached session or MFA claim can change the prompts you see.

The policy does not appear in the expected sign-in
Confirm the event time, tenant, resource, client and account. Recheck policy state and group membership, then try a fresh interactive sign-in.
The user signs in successfully but the policy says more user action is required
Check that the result belongs to the report-only policy. Compare its grant evaluation with the overall sign-in result and review the other policies applied to that event.
A pilot user cannot complete sign-in after enforcement
Use the event details and request identifiers to identify the blocking control. Check MFA registration and other applicable policies. If impact is unexpected, follow the rollback criteria before expanding scope.
RS / INTERACTIVE SHELL
RYANSHENTON.DEV // TERMINAL MODE Explore the published knowledge index. Type 'help' to get started.
visitor@rs:~$

Search the site

Search titles, topics or document text. Use ↑ and ↓ to move through results.