What Is User Provisioning? A Practical Guide for Hybrid Microsoft Environments

An employee moves from support into finance. On day one, they get the finance permissions they need — shared drives, the billing system, the finance distribution list. What they keep, without anyone deciding to keep it, is every support permission they already had: the ticketing queue, the support shared mailbox, a handful of admin groups nobody remembers assigning. Nobody notices, because nothing breaks. Three years later, that account can touch two departments’ worth of sensitive data, and no single person could tell you why.

That is not a user problem. It is a provisioning problem.

User provisioning is the process of creating, updating, and removing a person’s accounts and access across every system they need to do their job. Get it right and new hires are productive on day one. Get it wrong — as most organizations quietly do — and access accumulates for years with no one accountable for taking it back.

This article covers what provisioning actually is and where it stops, the three decisions that determine whether it holds up as you scale, and how to tell whether your own environment already has the gaps described below. Organizations that want a structured way to fix this outright can start with Identity Governance & Lifecycle Automation; everyone else, keep reading.

Know Exactly What User Provisioning Covers — And What It Doesn't

At its simplest, user provisioning means creating, updating, and removing a user account and its permissions across the systems a person needs to do their job. The important word there is permissions, not account. Creating the account is the easy part — a username, a password, a mailbox. The entitlements attached to it are where provisioning actually fails: group memberships, application licenses, shared drive access, distribution lists. An account with no entitlements is useless; an account with the wrong entitlements is a liability.

It’s also worth being precise about what provisioning is not, since a lot of content on this topic blurs the line. Provisioning creates the user identity and grants permissions. Authentication proves the person using that identity is who they claim to be. Authorization decides what that identity can reach at the moment of access. These are three separate systems of control, and conflating them is exactly why provisioning tends to get under-scoped as an account-creation task instead of the ongoing access-management discipline it actually is.

Deprovisioning is the same process run in reverse, and it’s the half most organizations execute badly. It gets its own full treatment later in this article, because it deserves one.

Follow One User Account Through Its Entire Lifecycle

The clearest way to understand provisioning is to follow one person through it — the pattern identity teams call joiner, mover, leaver (JML).

Joiner. A new employee’s role, department, location, and employment type determine their birthright access — the baseline permissions granted automatically the moment their account is created. Done well, this is the visible win of good provisioning: day-one productivity instead of a week of access-request tickets. The hidden risk is that birthright access rules get defined once, at rollout, and then almost never revisited as the organization’s roles and systems change underneath them.

Mover. This is where real risk lives, and it deserves more attention than either of the other two stages. A role change reliably adds access and almost never removes it — the pattern identity teams call privilege creep. It goes unnoticed for a structural reason: departures are discrete, visible events that trigger an offboarding process, but role changes are gradual and usually have no owner at all. Nobody is explicitly accountable for stripping the support permissions once someone moves into finance. That’s not a careless employee holding onto access — it’s a provisioning failure, and it compounds. An employee with five years and three role changes behind them typically carries permissions from all three roles, not just the current one.

Leaver. Offboarding usually means revoke, disable, and eventually remove — but disabled does not mean deprovisioned. Group memberships survive a disabled account. Mailbox delegation and shared access survive it. Application entitlements in downstream SaaS tools often survive entirely, quietly active in systems nobody thought to check. Contractors and third parties are the hardest version of this problem, because there’s frequently no HR record to trigger the process in the first place — they can leave a project with no system anywhere registering that fact.

Pick the Provisioning Model That Matches Your Reality

Partner with Microsoft experts you can trust

If it’s time to take that first step toward leveling up your organization’s security, get in touch with Ravenswood to start the conversation. 

There isn’t one correct way to provision access — there are several models, and most environments run a blend of them, deliberately or not.

Manual provisioning — tickets, spreadsheets, and tribal knowledge — is where a large share of mid-market organizations still operate, and there’s no shame in that; it’s a starting point, not a failure. Its failure modes are predictable: inconsistency between who handles a request, delay in getting people access, and no reliable audit trail when someone later asks who approved what.

Automated user provisioning ties account creation and access assignment to an event in an authoritative system, usually HR — a hire, a transfer, a termination — that triggers everything downstream. It only works as well as the trigger source feeding it, a problem this article comes back to.

Role-based access control (RBAC) assigns access by role or attribute instead of individual request. The prerequisite almost nobody mentions is that RBAC requires accurate, well-maintained role definitions before it works — and most organizations don’t have them. Attempting RBAC without that foundation produces roles that fit nobody cleanly.

Just-in-time provisioning creates the account at first sign-in rather than in advance — a strong fit for contractors, seasonal staff, and external collaborators who need access briefly and predictably.

Self-service with approval workflows lets users request access, with a designated owner approving before the system executes the grant. It only works when that approver genuinely understands what they’re approving — a rubber-stamped workflow is functionally no approval at all.

Almost no real environment runs a single model end to end. The goal isn’t picking one — it’s mixing them deliberately, matched to each population, instead of the accidental sprawl that happens when nobody decided which model applies where.

See Which Systems Do the Work in a Microsoft Environment

In most modern Microsoft estates, Microsoft Entra ID is the identity provider and the hub the provisioning service runs through, but it rarely acts alone.

Inbound, HR-driven provisioning typically starts with a platform like Workday or SuccessFactors. A hire, transfer, or termination event becomes the trigger, and the HR record becomes the starting point of the entire lifecycle — everything downstream depends on that record being accurate and timely.

Outbound, Entra ID pushes account and access changes to applications using SCIM — an open standard that lets identity systems and applications exchange user account information automatically. What it buys you is simple but valuable: one change in the source system propagates everywhere it needs to, instead of someone manually updating each application.

Directory synchronization between on-premises AD DS and Entra ID is the plumbing that makes any of this possible in a hybrid estate. On top of it sits governance — access packages and access reviews — the layer that keeps provisioning honest over time rather than accurate only on the day an account was created.

One line worth adding: legacy and on-premises systems that cloud-native provisioning simply cannot reach still need a deliberate plan, not an assumption that they’re covered.

Decide Where Your Source of Truth Lives Before You Automate Anything

Every provisioning system depends on an authoritative source — the system that owns the fact that a person exists, what their role is, and when that changes. Usually that’s HR. Sometimes it’s AD DS. In practice, it’s often neither, because nobody has actually decided.

This is the hybrid trap: many organizations assume Entra ID is authoritative simply because it’s the modern, cloud-facing piece of the stack, when on-premises AD DS is still where identities are actually born and modified day to day. The gap between that assumption and the reality is where provisioning projects quietly fail, long after the tooling has been purchased and configured correctly.

The consequence is direct. Automating on top of an unresolved source-of-truth question doesn’t fix bad data — it replicates it faster, across more systems than before. A useful diagnostic question for any environment: if a person’s title changes, which system finds out first, and what actually happens next?

Applications that will never support modern provisioning standards need a documented, deliberate answer too — not an assumption that someone downstream is quietly handling it. At a high level, that’s usually a scoped manual process with a named owner, or a custom integration. Both are decisions made on purpose; neither should be a default nobody chose.

Most failed provisioning projects fail on this question, not on tooling. It’s worth working through source-of-truth clarity before evaluating a single product — and it’s the starting point when identity governance engagements begin.

Keep Privileged Accounts Out of Your Standard Provisioning Flow

Automating birthright access is good practice. Automating privileged access the same way is how you industrialize a breach.

Privileged accounts need their own lifecycle for reasons that don’t apply to standard users: a separate approval path built around a human decision rather than an attribute match, credentials separated from the person’s day-to-day account, access that’s time-bound instead of standing indefinitely, and a review cadence far tighter than a standard account requires.

This is where the concept of tiered access becomes relevant, even for a reader who isn’t an identity architect: the most sensitive administrative accounts belong in their own tier, with their own controls, isolated from standard user provisioning entirely. That’s worth one clear paragraph here rather than a full architectural explainer — the point is that a provisioning project that treats privileged accounts the same as everyone else’s has already made its most consequential mistake.

This is also the point where a provisioning project stops being an IT efficiency exercise and becomes a security architecture decision — and where the conversation benefits from someone who has designed one of these before.

Don't Let Service Accounts Slip Outside the Process

Non-human identities — service accounts, application identities, automation credentials — are usually created ad hoc, outside any formal process, and never enter the lifecycle at all.

The consequence is predictable: no owner, no expiry, no review. These accounts are frequently still active, and often still privileged, five years after the project that created them ended, and they’re commonly over-permissioned at creation simply “to make it work” during setup.

The fix is one concrete rule, not a program: assign a named human owner and a review date at the moment any service account is created, and treat that as non-negotiable — including for accounts meant to be temporary.

Find the Provisioning Gaps Already Sitting in Your Environment

At this point you know what good provisioning looks like. Here’s how to check your own environment against it today, not hypothetically:

  • Accounts that have never once logged in.
  • Disabled accounts that still hold group membership.
  • Users sitting in privileged groups long after a documented role change.
  • Contractor and vendor accounts with no expiry date set.
  • Nested group structures nobody on the current team can explain or unwind.
  • Service accounts with no identifiable owner.

None of these are rare or hypothetical — they’re the routine findings of nearly every identity assessment Ravenswood runs. If any of them sound familiar, a free assessment is a low-commitment way to see exactly where your environment stands — a diagnostic conversation, not a sales pitch.

Turn Provisioning Into Evidence You Can Hand an Auditor

Provisioning isn’t just an operational concern — it connects directly to the compliance frameworks that create budget pressure: access recertification under SOX, HIPAA’s minimum-necessary standard, and CMMC 2.0’s access control and identification/authentication requirements for defense-industrial-base and federal organizations.

A practical budget reality worth naming plainly: provisioning work tends to get funded when it shows up as an audit finding, and rarely when it’s pitched purely as an efficiency gain. That framing is worth borrowing for an internal conversation, whatever the actual audit trigger turns out to be.

Build Provisioning That Holds Up as You Grow

Three ideas are worth carrying forward from all of this: settle your source of truth before you automate anything, treat the mover stage as the real risk rather than the leaver stage, and keep privileged accounts on a separate, governed path from everyone else.

When your organization is ready to design this properly, identity governance is where that work starts. And if you’re not yet sure where you stand, a free assessment is a good next step either way. We’re glad to help you figure out what “good” looks like for your environment.

[RELEVANT BLOG CONTENT]