What Is Conditional Access and How Do You Plan It?

How does Conditional Access evaluate an access request, and which license does it need? Starter policies, report-only mode and the rollout order with emergency accounts.
In this article5 sections

A mistake in Conditional Access either leaves a security gap or stops the whole company from signing in at nine in the morning. Anything in between is rare. Conditional Access is the policy engine that determines Microsoft Entra ID’s decision on a sign-in request based on the user, application, device, location and risk. In Microsoft’s definition, it’s at the center of the policy decision point in the Zero Trust approach. Below we explain the engine’s logic, a safe starter policy set and how to roll that set out without locking yourself out of the tenant.

How does it work?

Each policy has two parts: assignments (who accesses which application under which condition) and access controls (block, require MFA, require a compliant device, limit session lifetime). During sign-in, all applicable policies are evaluated together; if a policy blocks access, the block takes precedence.

The point often misunderstood here is this: policies aren’t processed in order like firewall rules, and there’s no “first match wins” logic. All the requirements of every policy in scope are applied together. If one policy requires MFA and another a compliant device, the user has to satisfy both. If one policy blocks, it doesn’t matter what the others say.

Conditional Access evaluates the relevant access request against policy conditions after the initial authentication. This applies to sign-ins with passwordless methods such as passkeys or Windows Hello for Business as much as to those with passwords. It doesn’t stop password spraying attempts itself; after a successful password guess, the policy decides whether access is granted.

Signal Example
User / group / role All users, administrator roles, guests
Target resource Office 365, Azure management, a specific enterprise application
Instrument Intune-compliant device, hybrid-joined device, platform
You are here Country, named IP ranges
Client app Browser, modern client, legacy authentication
Risk (P2) Sign-in risk, user risk

Don’t rely too much on the location signal. Defining the office IP as a “trusted location” and exempting sessions from there from MFA is a common choice; but a compromised computer on the office network also comes from the same IP. Location works more reliably as a blocking condition (e.g. countries where you don’t do business).

License requirement

Conditional Access requires Microsoft Entra ID P1; bundles such as Microsoft 365 Business Premium and E3 include P1. Risk-based conditions require P2. Tenants without P1 should use security defaults for basic protection.

Security defaults and Conditional Access don’t work together. You need to disable security defaults before turning on the first Conditional Access policy. This switchover creates an unprotected gap. So prepare your policies in report-only mode in advance and turn on the MFA policies the moment you disable the defaults; reduce the gap from hours to minutes.

Starter policy set

  1. MFA for administrator roles: Strong authentication for all directory roles.
  2. MFA for all users: Excluding emergency accounts.
  3. Block legacy authentication: Shuts down protocols that bypass MFA.
  4. MFA for Azure management: Portal, CLI and PowerShell access.
  5. Compliant device for corporate data: An Intune compliance or app protection policy condition.
  6. Block access from countries you don’t operate in: According to business needs, with an exception process.

In some tenants, Microsoft also rolls out such basic controls as “Microsoft-managed policies”; they should be reviewed so they don’t conflict with your existing policies.

Why does the order matter?

The administrator policy comes first because few people are affected while risk reduction is highest. Before blocking legacy authentication, look at the “client app” column in the sign-in logs: you’ll find a scanner-printer still using IMAP, POP or SMTP AUTH, an old ERP integration or a very old mail client. If blocking is turned on before these are identified and moved to modern authentication or an alternative method, work stops.

The compliant device policy comes last. If it’s turned on while there are devices whose Intune enrollment isn’t complete, a significant share of users lose access. Monitor the device enrollment rate in the reports first.

When MFA isn’t enough: authentication strength and session controls

Three things are separate: a user registering an MFA method, a policy requiring MFA for access, and an authentication strength that narrows the acceptable methods. The “require MFA” control accepts every registered method available to the user, including SMS and voice calls. That isn’t enough for administrator roles. With the “authentication strength” option in the access control, you can require phishing-resistant methods (FIDO2 security key, Windows Hello for Business, certificate-based authentication) from administrators. A common picture: only SMS is registered on the administrator account.

Session controls determine how long the decision stays valid. Shortening sign-in frequency for administrator portals and browser access from unmanaged devices reduces the impact of long-lived sessions being stolen. Don’t force users to re-authenticate every hour; MFA fatigue is also a risk.

Safe rollout: report-only mode

New policies should first run in report-only mode. In this mode the policy isn’t enforced, but the sign-in logs show “what would have happened if it had been applied”. After a few days of observation, unexpected effects are fixed and the policy is turned on.

Turning on report-only mode and forgetting it is also a mistake. Set the observation period up front; it should cover at least one business cycle so you also see infrequent processes such as month-end close and payroll. Group the sign-in log entries with the result “Report-only: would have failed” by user and application. When the list gets short, turn the policy on first for a pilot group, then for everyone.

The “What If” tool in the Entra admin center is also useful: it shows which policies would apply for a given combination of user, application and location without touching live traffic. After every new policy, run this tool with the emergency account and verify the account isn’t caught by any policy.

According to Microsoft’s guidance, at least two emergency accounts are kept, and these accounts are excluded from policies that block or restrict sign-in; exclusion isn’t needed for policies in report-only mode. Because the MFA requirement on admin portals also covers these accounts, passkeys (FIDO2) or certificate-based authentication are configured for them.

Common mistakes

  • Turning on “block all users”-type policies without excluding emergency accounts, and getting locked out of the tenant.
  • Growing exclusion groups and forgetting them; exclusions should be reviewed regularly.
  • Not accounting for service accounts and automations.
  • Not standardizing policy names; readable naming (e.g. CA01-Admins-MFA) makes management easier.

Exclusion groups need extra attention. An administrator left out of MFA “temporarily” is still there six months later. Assign owners to exclusion groups, tie membership to periodic access reviews and write the reason for the exclusion in the group description.

For service accounts, the right solution is usually not to exempt them from user policies but, where possible, to move them to managed identities or workload identities. For accounts that can’t be moved, limit the exception to a specific IP range and specific application. A broad exemption is a ready-made door for an attacker.

A final check: record every policy change. Who changed which policy and when is visible in the audit logs; define alerts for these changes.

For Conditional Access design and phased migration, see our Microsoft 365 Security and Identity and Access Management services, and for the long-term approach, our Zero Trust Security Architecture page.

Support

Contact us for support or a discovery session on this topic.

Contact form
[email protected]

Eretek Cybersecurity

Firewall, VPN, endpoint security and threat prevention. We reduce the attack surface and measure and report the risks.

Knowledge Center

Other articles