Your MFA policy may be enabled, your sign-in dashboard may look healthy, and important accounts may still be outside the protections you think they have.
By AzureWise | Microsoft Entra ID security
Microsoft Entra Conditional Access is one of the most effective controls available to identity administrators. But a policy that exists in the portal is not proof that the right identities, applications, and sign-in scenarios are protected.
The difficult question is not “Do we have Conditional Access?” It is “Can we demonstrate that the policies we intended to enforce actually apply to the access paths that matter?”
A tenant can require MFA for employees and still have overbroad exclusions. It can require MFA for administrators without requiring phishing-resistant methods. It can have carefully designed user policies that never apply to a service principal. And it can have a promising new policy that has remained in report-only mode for months.
Here are five gaps worth checking in your next Microsoft Entra ID security review, along with concrete ways to investigate them.
1. Exclusions have become an invisible access path
Exclusions are often introduced for a legitimate reason: a temporary application issue, a rollout exception, or an emergency-access account. The problem starts when temporary exceptions become permanent and no one can explain who owns them.
For an individual Conditional Access policy, an explicit exclusion takes precedence over an inclusion. An “All users” assignment therefore does not automatically mean all users are covered if exception groups or accounts are excluded.
What to audit:
- Open each high-impact policy and inspect excluded users, groups, roles, and applications.
- Identify the real membership of every exclusion group, including who can change it.
- Record a business owner, justification, review date, and expiration or remediation plan for non-emergency exceptions.
- Use Microsoft Entra access reviews where appropriate to periodically revalidate excluded group membership.
Important distinction: Emergency or break-glass accounts are a deliberate exception. Microsoft recommends retaining emergency access accounts outside restrictive Conditional Access policies to reduce lockout risk, while protecting them through separate strong authentication, tightly controlled access, testing, and alerting. Do not “fix” exclusion sprawl by blindly removing those safeguards.
Evidence to retain: A dated policy export, exclusion membership list, approved exceptions, and the outcome of the latest review.
2. “Require MFA” is being mistaken for phishing-resistant MFA
MFA is a crucial baseline, but not every MFA method offers the same protection against phishing. A general MFA grant control does not, by itself, establish that administrators must use phishing-resistant methods.
Microsoft Entra Conditional Access supports authentication strengths, including a built-in Phishing-resistant MFA strength. That gives teams a more precise control for high-value users and resources, rather than relying on a broad “MFA required” label.
What to audit:
- Review Conditional Access requirements for privileged directory roles and sensitive applications.
- Determine whether the grant control is general MFA or an explicitly configured authentication strength.
- Confirm that enrolled methods and client scenarios can satisfy the proposed strength before enforcement.
- Test representative administrative sign-ins using the Conditional Access What If tool and report-only evaluation.
A useful question for the review meeting: “Which exact authentication methods are permitted when our most privileged administrators sign in?” If the answer is unclear, the policy name alone is not sufficient evidence.
Evidence to retain: Policy assignments, configured authentication strengths, test scenarios, and rollout results.
3. Workload identities are outside the user-policy safety net
This is a particularly easy gap to miss because the application may function perfectly while its identity risk remains unexamined.
Traditional user-scoped Conditional Access policies do not apply to service principal sign-ins. Microsoft offers Conditional Access for workload identities for supported single-tenant service principals, subject to licensing and feature limitations. These policies can block token requests based on supported location or risk conditions. They do not work like a user MFA challenge: service principals cannot complete interactive MFA.
The scope matters. Microsoft documents that these policies do not cover managed identities or multitenant third-party SaaS applications in the same way. Assigning a service principal to a group covered by a user-scoped policy does not make that policy enforceable for the service principal.
What to audit:
- Inventory application registrations and enterprise applications, prioritizing highly privileged service principals.
- Review credential types, secret and certificate lifetimes, permissions, owners, and recent sign-in activity.
- For eligible service principals, assess whether workload identity Conditional Access and risk monitoring are appropriate.
- Prefer managed identities where the architecture supports them, while evaluating their permissions and monitoring separately.
- Look at service principal sign-in logs—not only interactive user sign-ins.
Evidence to retain: A service principal inventory, risk ranking, credential ownership, monitoring coverage, and documented control decisions.
4. Report-only policies were never promoted to enforcement
Report-only mode is valuable because it lets teams evaluate the likely effect of many Conditional Access controls before they interrupt production access. The trap is treating report-only results as if the policy were already protecting users.
In report-only mode, a policy is evaluated but is not enforced. A result such as reportOnlyFailure means that the simulated policy requirements were not met; it does not mean that this report-only policy blocked the sign-in. Another enforced policy may still have blocked it.
What to audit:
- List all policies in Report-only and Off states and identify why they are not enabled.
- For each report-only policy, review representative sign-ins and determine the expected impact.
- Assign an owner and a decision date for enabling, redesigning, or retiring each policy.
- Roll out changes in stages and retain a tested rollback approach.
Do not change an already enforced policy from On to Report-only simply to test it: doing so removes its enforcement. Microsoft recommends evaluating proposed policies alongside existing enforced policies.
Evidence to retain: Test results, expected exceptions, approval to enable, actual enablement date, and post-deployment monitoring.
5. Policy presence is measured, but policy effectiveness is not
It is easy to produce a list of Conditional Access policies. It is harder—and much more useful—to answer whether the controls are actually applying to important sign-ins.
The Entra sign-in logs and Conditional Access evaluation details provide a place to start. Review both policies that applied and policies that did not apply. A “not applied” result is not necessarily a security defect: the user, resource, and conditions may legitimately fall outside that policy. The task is to identify unexpected nonapplication for high-risk scenarios.
A simple validation workflow:
- Select a critical scenario: privileged administrator, external collaborator, sensitive application, or an eligible service principal.
- Define the expected control: phishing-resistant MFA, access block, or another applicable requirement.
- Simulate the scenario with What If where supported, then inspect real sign-in evidence.
- Review whether each expected policy applied and, if not, why.
- Track unexpected outcomes to remediation and rerun the test after changes.
For environments forwarding sign-in logs to Log Analytics, this KQL query can help surface per-policy evaluation results for further investigation:
SigninLogs
| where TimeGenerated >= ago(7d)
| mv-expand Policy = ConditionalAccessPolicies
| extend PolicyName = tostring(Policy.displayName),
PolicyResult = tostring(Policy.result)
| summarize SignIns = count(),
Users = dcount(UserPrincipalName)
by PolicyName, PolicyResult
| order by SignIns desc
This is a triage view, not a coverage score. It requires the SigninLogs table to be available and populated in the selected Log Analytics workspace. It counts policy-evaluation records, not unique sign-in events across all policies; a single sign-in can produce multiple policy rows. A result of notApplied or reportOnlyNotApplied requires scenario and policy-scope context before it can be called a gap. Policies absent from the logs cannot be judged from this query alone.
Evidence to retain: Expected-control test cases, sign-in evaluation results, open gaps, and closure verification.
The 30-minute Conditional Access reality check
You do not need to redesign every policy in one sitting. Start with a focused review:
| Time | Check | Question to answer | | --- | --- | --- | | 0–5 min | Policy inventory | Which critical policies are On, Report-only, or Off? | | 5–10 min | Exclusions | Who is excluded, and can we justify every exception? | | 10–15 min | Privileged access | Are administrative sign-ins required to use phishing-resistant MFA? | | 15–20 min | Workload identities | Which high-privilege service principals have appropriate controls and monitoring? | | 20–30 min | Evidence | Can we demonstrate expected policy evaluation for key sign-in scenarios? |
This is a first-pass screening, not a complete Conditional Access assessment. The point is to turn five vague assumptions into five testable questions.
The takeaway: validate outcomes, not policy counts
A mature Microsoft Entra security program should be able to show not just what is configured, but who is covered, what exceptions exist, how policy decisions are evaluated, and who investigates the gaps.
Conditional Access is not a set-and-forget control. Application inventories change. Privileges change. Exceptions accumulate. Workload identities multiply. And a policy that looked sensible six months ago may no longer match the tenant it is supposed to protect.
The next time someone asks whether your Microsoft Entra tenant is protected by Conditional Access, do not lead with the number of policies you have. Lead with the evidence that the policies are working where they matter.
About AzureWise: AzureWise publishes practical Microsoft Entra ID security guidance and tools for identity and cloud security teams. Explore the AzureWise Identity Security Assessment for additional visibility into identity risks, privileged activity, and sign-in patterns. The assessment complements—but does not replace—a full Conditional Access policy review.
Microsoft documentation and further reading
- Build Conditional Access policies
- Conditional Access: users, groups, agents, and workload identities
- Manage Conditional Access exclusions with access reviews
- Require phishing-resistant MFA for administrator roles
- Conditional Access for workload identities
- Analyze Conditional Access policy impact
- Troubleshoot with the What If tool
- Manage emergency access accounts
- Conditional Access insights and reporting
Want the Azure identity assessment already built?
AzureWise Identity Security Assessment turns Microsoft Entra ID telemetry into a weighted Security Rating, detailed findings, and prioritized remediation guidance inside Azure Monitor.
Get AzureWise — $99