Count the administrators first. Then go and count the service accounts, the supplier logins, and every local account sitting on every server you own.
The point of the exercise is to find every route to elevated capability, and the directory holds only some of them. What actually belongs on the list is the service account with no owner, the supplier account nobody has reviewed since it was issued, the local account whose password has never changed, and the automation credential living inside a script on a file share.

- Every pathNot only directory administrators
- CIS 5, 6 and 8Accounts, access control, audit logging
- Service accountsThe population nobody has an owner for
- Break-glassVerified by signing in, not by reading
Seven distinct privileged populations exist. In most organizations, only the first one has ever been counted.
Directory administrators, which everybody counts
Global administrators, domain administrators, and everything in the tiers underneath. This is the one list a business can put together when asked, and it almost always turns out bigger than whoever compiled it expected once the nested groups are flattened out. It is also the only list most organizations ever review, which is precisely why an audit that stopped here would be worthless.
Service accounts, which almost nobody owns
Every application, backup job, integration and migration leaves accounts behind. They typically hold more privilege than the work ever needed and carry passwords set on the day they were created. Counting them is not the useful question. The useful question is how many have a named human who could tell you what breaks the moment that password is rotated.
Local accounts on servers and appliances
Local administrator on the Windows servers. Root and sudo on the Linux boxes. The built-in accounts on switches, storage arrays, hypervisors, backup appliances and the building management system. None of them live in the directory, which means no directory-based review has ever seen them, and in a great many estates they all share one password.
Vendor and support access
An account handed to a supplier for support. A remote access tool installed for a project that finished. Standing credentials given to an integrator several years and two staff changes ago. This population is rarely documented, usually still live, and every entry in it is a path into your systems that runs through a company whose security controls you have no say over.
Cloud and platform privilege beyond the directory
Subscription owners, role assignments made at resource level, who can open a key vault, application registrations carrying high consent, and the service principals and managed identities holding more permission than their workload has ever used. Your directory administrator list tells you nothing whatsoever about who is able to delete a production resource group, and that has become the more consequential question of the two.
Standing versus elevated on demand
Counting who holds privilege is only half of it. The audit also records whether they hold it permanently or activate it when there is a reason to. Permanent assignment remains the norm almost everywhere, and changing it moves the risk position further than any other single finding, because a standing target becomes a time-limited one.
Whether any of it is logged and looked at
Control 8 covers audit log management, and privileged activity is where that control earns its keep. Three questions follow from it. Are privileged sign-ins and actions logged at all. Are those logs kept long enough that an investigation is possible. And would anyone actually notice an unexpected one. It is the third question that organizations fail.
The biggest privileged population in your estate is service accounts, and it is the one nobody governs.
Ask for a list of administrators and any business can produce one. Ask for a list of service accounts with a named owner beside each entry and almost none can.
- Each one gets created for a job, granted whatever privilege made the job start working rather than the privilege the job actually needed, and then left alone permanently, because touching it risks breaking whatever it quietly runs.
- Their passwords usually date from the day they were created, and the reason is always the same: nobody can say with any confidence what stops working if it changes. It is exactly the asymmetry that keeps obsolete firewall rules in place, transplanted onto credentials.
- They are simultaneously the accounts least likely to carry multifactor authentication, most likely to be carved out of conditional access, and most likely to be written in plain text into a script, a scheduled task or a configuration file that a great many people can read.
- What breaks this cycle is not a number. It is a list carrying a named owner against every account, a plain statement of what each one does, and a rotation plan for those where the owner can confirm the blast radius. Building that list accounts for most of the effort in the engagement and essentially all of the value.
Four things that separate this from handing you a list of administrators.
We inventory service accounts and chase every owner
In most businesses this is both the largest privileged population and the one under the least governance. A count achieves nothing. What matters is a list carrying a named owner, a stated purpose, and an honest assessment of what a rotation would take down. You cannot query a directory for that. It comes from sitting down with the application teams.
Our attention goes outside the directory, which is where most of the estate actually is
Local accounts on servers. Linux sudoers files. Credentials on network devices. Hypervisor and storage consoles. Backup systems. Database logins. Building management. A directory review sees none of them, every one of them grants privilege, and several grant more than any directory role could, for the simple reason that they sit underneath the directory rather than inside it.
We test the break-glass accounts by using them
The guidance on emergency access accounts leaves little room for interpretation. At least two of them. Cloud-only, on the onmicrosoft.com domain. Phishing-resistant authentication that differs from what the normal administrative accounts use. Permanently active in Privileged Identity Management rather than eligible. Excluded from any Conditional Access policy that could block them. Alerted on. And validated no less often than every ninety days. We verify that last point by actually signing in, not by reading a configuration screen and assuming.
The question we ask is whether anyone would notice, not merely whether it was recorded
Control 8 is audit log management, and privileged activity is where it earns its place. Plenty of organizations log privileged sign-ins. Rather fewer keep those logs long enough for an investigation to be possible. Almost none have somebody who would spot an unexpected privileged sign-in at three in the morning, and that last point is what separates an actual control from a filing cabinet.
Six US situations where a privileged access audit changes the picture.
A regulated firm asked to evidence privileged access control
A SOC 2 examination, a HIPAA security assessment, NYDFS Part 500 and the FTC Safeguards Rule all want the same four things from privileged access: that it is restricted, justified, monitored and reviewed on a schedule. Respond with a directory administrator list and you have addressed a fraction of the privileged estate. The full audit answers for every population, and that is what each of those questions is really asking even where the wording never says so.
A group that has grown by acquisition
Every acquisition arrives with a directory of its own, its own service accounts, its own supplier relationships and its own habits around local accounts. Integration then usually bolts on cross-domain trust or synchronization without anybody auditing what privilege that quietly grants. Nobody assembles the combined picture unless somebody is told to, and a diligence team on the buying side will eventually ask for exactly that picture.
An organization with heavy supplier and integrator involvement
Supplier accounts, support access and remote tools installed for projects pile up steadily and get removed almost never. Every one of them is privilege inside your environment held by a company whose security controls you have no visibility into, which makes each entry a security question and a third-party risk question at once.
An operator with technology outside the IT estate
Building management, door access control, plant systems and facilities technology all sit on the network, all hold privilege over things with physical consequences, and almost none of them ever appear in an identity review. Default credentials are common, for a straightforward organizational reason: IT does not regard them as its responsibility and facilities does not regard them as a security asset.
An organization that has just had an identity incident
Once an account is compromised the first question is what it could reach. The second and larger question is what else in the estate has comparable reach, and only a full privileged inventory answers that. Most organizations commission this work in the weeks after an incident, when it costs considerably more than it would have beforehand.
A business implementing just-in-time privileged access
Before standing privilege can become on-demand privilege, somebody has to know what standing privilege exists, including the parts that cannot move at all because they are service accounts or appliance logins. The audit has to come first, and what it generally reveals is that the human administrators were the straightforward half of the problem.
How US organizations understand their privileged access.
| Feature | Full privileged access audit | Directory administrators reviewed | No privileged access review |
|---|---|---|---|
Directory administrators known | Yes | Yes | No |
Service accounts inventoried with owners | Yes | No | No |
Local and appliance accounts covered | Yes | No | No |
Vendor access documented | Yes | No | No |
Cloud resource privilege included | Yes | Rarely | No |
Standing versus on-demand recorded | Yes | No | No |
Credential age known | Yes | Partly | No |
Break-glass accounts tested | Yes | No | Not applicable |
Privileged activity monitored | Yes | Logged only | No |
Answer to who could do the most damage | Yes | Partial | No |
Nine questions, asked of every privileged account we find.
Question
Who is the named owner?
- Why it matters
- Without a named owner an account can never be reviewed, altered or switched off
Question
What is it for?
- Why it matters
- Purpose determines whether the privilege level is proportionate
Question
Does the privilege match the purpose?
- Why it matters
- Most over-privilege is historic rather than deliberate
Question
Is it standing, or activated on demand?
- Why it matters
- Standing privilege is a persistent target
Question
How does it authenticate?
- Why it matters
- Phishing-resistant, multifactor, a password on its own, or a key sitting in a file
Question
When did the credential last change?
- Why it matters
- And does anybody know what rotation would break
Question
When was it last used?
- Why it matters
- Unused privileged accounts are the cheapest thing to remove
Question
Is its activity logged?
- Why it matters
- Control 8 covers audit log management, and privileged activity is where it counts
Question
Would anyone notice unexpected use?
- Why it matters
- Logging without review is storage, not detection
Five steps, of which the second consumes more time than the other four combined.
- 1
Enumerate every privileged path, not only the directory
We enumerate directory roles with every nested group flattened out, local administrators on servers, root and sudoers on Linux, database logins, accounts on network devices, hypervisor and storage consoles, backup systems, cloud subscription and resource roles, application registrations and service principals, plus whatever facilities or operational technology falls inside the agreed scope.
- 2
Attribute ownership, especially for service accounts
We sit down with the application and platform teams and put a named owner and a stated purpose beside every account. It is the slowest part of the engagement, and it is what decides whether you receive a list of decisions or merely a census. Any account that comes out of this step still without an owner is a finding in its own right.
- 3
Assess proportionality, standing privilege and credential hygiene
Five things per account: whether the privilege actually matches the stated purpose, whether it is standing or activated on demand, how it authenticates, when its credential last changed, and when anyone last used it. That final field matters more than it sounds, because unused privileged accounts are both the cheapest remediation available and, reliably, the largest single category in the report.
- 4
Test break-glass and check monitoring
We sign in to every emergency access account rather than reading its configuration and hoping, and we confirm the alert actually fires. From there we check whether privileged sign-ins and actions are logged, whether those logs survive long enough to investigate with, and whether anyone reviewing them would spot something that did not belong. That last one is where organizations usually come unstuck.
- 5
Deliver a prioritized remediation plan with owners
Unused accounts get removed. Over-privileged ones get scoped down. Standing privilege becomes on-demand wherever the account type permits it. Credentials rotate wherever an owner can confirm what that affects. Supplier access is documented and given an expiry. Every item carries a named owner, because a privileged access report without owners against its findings changes precisely nothing.
What US organizations ask about privileged access audits.
Fifteen sources of privileged access to include in scope.
The obvious ones
- Directory administrative rolesWith nested groups expanded.
- Server local administratorsOutside the directory, outside most reviews.
- Linux root and sudoIncluding sudoers entries.
- Database administrative accountsFrequently separate from the directory.
- Network device accountsOften shared, often unchanged.
The ones that get missed
- Service accounts across all platformsThe largest population.
- Application registrations and service principalsConsent can exceed any human role.
- Cloud subscription and resource rolesNot visible in a directory role list.
- Hypervisor and storage consolesPrivileged over everything they host.
- Backup system accountsAble to read or destroy everything.
The ones nobody mentions
- Vendor and integrator accountsUsually standing, usually undocumented.
- Remote access tools installed for projectsFrequently still running.
- Break-glass accountsVerified by signing in, not by reading.
- Building and facilities systemsPrivileged, networked, and unowned.
- Credentials in scripts and config filesMachine secrets scanning finds these.
The pages around this one.
Request a list of every service account you run, with a named owner beside each one.
Making that request costs you a minute, and no other question in this area tells you as much. In most businesses the list either does not exist at all or turns up with the owner column blank. Whichever answer you get is itself the argument for running the audit.
Related Services
Explore more solutions that work great with this service
Microsoft Entra Privileged Identity Management
Privileged Identity Management deployment for US organizations:
Learn moreEmergency Access Account Design
Break-glass emergency access account programs for US organizations:
Learn moreMicrosoft Entra Access Reviews
Entra access review programs for US organizations: entitlement
Learn moreMicrosoft Entra Workload Identities
Workload identity governance for US organizations: applications
Learn moreMicrosoft Defender for Identity Services
Identity threat detection for Active Directory and Entra
Learn more