The Sealed Envelope Test: Proving Emergency Admin Access Before the Day It Matters
Most UK organisations can point to an emergency administrator account. Far fewer can describe the last time it was opened, who opened it, what they did, and how the credentials were returned to custody. That gap matters. A break-glass account is not a control in itself; it is a procedure waiting to be tested, and an untested procedure is a hope with a login screen.
Define the failure before you choose the account
A useful rehearsal starts with the scenario, not the credential. The primary identity provider may be unavailable, conditional access may be misconfigured, the privileged role may be vacant, or the only administrator may be unreachable. Each scenario has a different trigger and a different approver. Write down the conditions that justify emergency access, who can declare them, and who must counter-sign. In a small UK business that may be the managing director and the IT lead; in a larger estate it may be a service continuity group. The point is to remove ambiguity before pressure arrives.
Custody is a rota, not a secret
Emergency credentials should have named custodians and named deputies. A sealed envelope in a safe, a hardware security key in a locked cabinet, and a documented handover are all stronger than a shared password in a team chat. The custodian list must be reviewed when people join, leave, or change role. If the credentials are held in a password manager, the emergency account should sit in a separate vault with access logging and a break-glass approval step. If they are physical, record the seal number and the date it was last checked. The question is not whether the credential is secret; it is whether you can prove who could have used it and when.
When the identity provider is the failure
This is where many rehearsals become uncomfortable. If the emergency account is federated to the same identity provider that has failed, it is not an emergency route. Microsoft 365 estates commonly use cloud-only emergency accounts that do not depend on federation, with permanent or highly privileged roles and strong authentication. Those accounts need a path that survives the outage: a hardware key held offline, a documented alternative sign-in URL, and a contact list that is not stored only in email.
Conditional access needs particular care. An emergency account that is excluded from every policy is a standing risk; one that is excluded from nothing may be blocked on the day. The workable pattern is a narrow, documented exclusion for the emergency account, protected by monitoring and alerting on every sign-in. During the rehearsal, confirm that the exclusion behaves as intended, that the MFA method is available, and that the account can perform the specific recovery action you need. Do not assume that a successful password reset proves emergency access. Prove the action.
Run the rehearsal under realistic pressure
Schedule the exercise in advance, but do not make it comfortable. Give the custodians a short window, a named approver, and a specific objective: sign in, confirm the role, perform a low-risk administrative action, and close the account down again. A low-risk action might be reviewing a role assignment, creating a temporary group, or confirming access to the recovery portal. The aim is to exercise the path, not to create an outage.
Capture the time from declaration to first successful sign-in, the time to complete the action, and the time to return the credential to custody. Record who approved the exercise and who performed each step. If the rehearsal is done in production, use a change record and a rollback note. If it is done in a test tenancy, be honest about what that does not prove: a test tenant rarely reproduces the conditional access policies, network restrictions, and human dependencies of the live estate.
Evidence for auditors and insurers
The evidence pack should be short enough to read and complete enough to defend. Include the scenario, the approver, the custodian, the timestamps, the sign-in logs, the action performed, and the resealing record. Screenshots are useful, but exported logs are stronger. Note any deviations and the remedial action. Retain the pack with the same discipline as a disaster recovery test. Insurers and auditors increasingly ask not whether an emergency account exists, but whether the organisation can show it has been exercised. A signed record of a rehearsal is more persuasive than a policy that has never been opened.
Three findings that almost always surface
The first is a conditional access mismatch. The emergency account is either excluded from so many policies that it is effectively unmonitored, or it is caught by a policy that nobody remembered. The rehearsal makes the gap visible.
The second is custody drift. The named custodian left six months ago, the envelope was resealed but not reissued, or the hardware key is in a drawer that no one can open. These are ordinary failures, and they are exactly why the exercise needs a rota rather than a legend.
The third is a hidden dependency. The procedure lives on the same intranet that is unavailable, the contact list is in the mailbox that cannot be reached, or the recovery instructions assume a mobile app that depends on the failed identity provider. Good emergency access is designed for the day when the normal route is gone.
Close the loop
A rehearsal is only finished when the findings have owners, dates, and a second test. Update the procedure, rotate anything that was exposed, confirm the custodians, and record the next exercise. The sealed envelope should not remain a symbol of preparedness. Open it, use it, close it, and write down what happened. That is how emergency access becomes evidence rather than assumption.