All Posts
Access ReviewsNovember 16, 2026 · 11 min read

What Is an Identity Governance Framework and Why It Matters

An identity governance framework is the operating model an organisation uses to decide who or what may access which resources, on what conditions, who owns each of those decisions, how access is granted and taken away...

What Is an Identity Governance Framework and Why It Matters

An identity governance framework is the operating model an organisation uses to decide who or what may access which resources, on what conditions, who owns each of those decisions, how access is granted and taken away, and how the organisation shows afterwards that its own rules were followed. Citadel Identity360 is built to run that model day to day, turning written policy into owners, workflows, reviews and records that cover employees, contractors, partners, service accounts and AI agents in one place.

It matters because access does not stay clean by itself. People change jobs, contractors finish, applications multiply and machine identities keep arriving. Without a repeatable way to decide, approvals drift, removals lag and nobody can say who holds what.

A framework is broader than software: policy, decision rights, ownership, review duties, evidence and the supporting technology. Astranova has already explained the parts of the tooling in what an access governance system is made of, and how board-level direction reaches identity teams in IT governance explained. This article sits between the two: the rules of the road for everyday access decisions.

Identity governance and IAM are not the same job

Identity governance asks whether access is appropriate, approved, owned, reviewable and evidenced. Identity and access management is the wider discipline; NIST describes its aim as making sure the right people and things have the right access to the right resources at the right time, which covers authentication and authorisation as well as governance.

Access can work perfectly and still be poorly governed. Passing every login check does not show that the access is justified, who approved it or whether it should still exist. IAM supplies the mechanics; governance tests whether the access can be defended.

In practice the boundaries overlap. IGA brings lifecycle administration together with access governance, including provisioning, policy enforcement and access reviews. The goal is knowing where each control lives and who answers for it.

Six design choices behind an identity governance framework

A workable framework is not a downloadable template. It is a set of decisions the organisation makes and keeps current.

Scope. Decide which identities are governed: staff, contractors, vendors, partners, privileged administrators, service accounts, APIs, cloud workloads and AI agents. Decide which applications, data and infrastructure fall inside the boundary. Our guide to external identities covers the non-employee population.

Sources of identity facts. Choose which systems or people establish employment status, organisational unit, role, sponsor and end date, because those facts drive access decisions. A source-of-record field still does not prove every attached permission is appropriate.

Accountability. Someone must own identity records, policies, approvals, applications, entitlements and access reviews. That can include IAM, security, HR, managers, application and data owners and compliance. In CSF 2.0, NIST asks that roles, responsibilities and authorities for cybersecurity risk management be established, communicated, understood and enforced.

Lifecycle triggers. Joiners, movers and leavers should never be handled as informal exceptions. Each event needs a defined trigger, approval path, responsible team and record, and non-human identities need the same from creation to decommissioning.

How access is expressed. Some organisations use roles, some attributes, some groups, some direct entitlements, and most use a mix. The test is fit with your applications and data.

Least privilege and conflicting duties. Set out how excess access, dormant access, privileged permissions and segregation-of-duties conflicts are handled across systems. The same NIST framework wants permissions, entitlements and authorisations defined in policy, then managed, enforced and reviewed on least-privilege and separation-of-duties principles. The framework should say when a conflict blocks access, when it needs extra approval and when it can be accepted for a period with a named owner and a remediation plan.

A decision-rights map for everyday access

Design choices become an operating model when every routine access decision has a named decider, a rule and a record. A short published map settles most arguments before they start:

  • Granting new access: the manager confirms business need, and the application or data owner confirms the entitlement is appropriate for the resource.
  • Owning an entitlement: one accountable owner per sensitive permission or role decides who should hold it and when it should be retired.
  • Accepting a conflict: a risk or control owner, not the requester or the requester’s manager, decides whether a segregation-of-duties conflict can stand and for how long.
  • Sponsoring outsiders: an internal sponsor answers for each contractor, partner or vendor identity and its end date.
  • Owning a machine or agent: a named person answers for every service account, API key and AI agent, including its purpose and retirement.
  • Removing access: the review owner or lifecycle event triggers removal, and the change is tracked from decision until it is fulfilled.

At control level, the AC-2 account management control in NIST SP 800-53 calls for assigning account managers, requiring approvals by defined personnel or roles for account requests, monitoring account use and reviewing accounts for compliance at a set frequency.

From one-off fixes to a repeatable access practice

Start with visibility. Build an initial inventory of identities, accounts, applications, entitlements and known owners. It will be incomplete at first, and the gaps are useful: they show where ownership is unclear, where accounts outlive departures and where approvals live in email.

Next, agree on policies and decision rights. Without a link to enterprise risk priorities, it becomes another tool that reports problems without changing how access is decided.

Then prioritise the highest-risk and highest-volume processes. Onboarding, role changes, departures, privileged access and critical applications are usually where failure hurts most.

After that, make lifecycle actions consistent. Joiner, mover and leaver events should create, change or remove access through the same governed path every time. Late or disputed events should surface, not disappear into manual work.

Finally, introduce structured reviews and remediation. A review campaign is not finished because it launched. The control outcome is whether unjustified access was actually removed or corrected.

What the framework delivers for risk, operations and people

Risk. Permission creep builds quietly as employees change roles, and inactive or departed identities that keep permissions create needless exposure. Governance finds and corrects those conditions; it lowers exposure rather than promising that no incident will happen.

Operations. Consistent lifecycle processes reduce reliance on memory and manual chasing. Automation helps only when ownership and data are sound; otherwise it moves inconsistency faster.

Employee experience. Clear rules, known owners and good workflows get people the right access consistently and expose avoidable service desk work.

Auditability. Recorded decisions, account changes, review outcomes and remediation give a defensible history of who had access, who authorised it and what happened when it changed. See where access evidence fits in IT compliance audits. Evidence supports an audit; it does not certify compliance by itself.

Bringing non-human principals inside the model

Governance scope now reaches well beyond people. The Cloud Security Alliance, in Defining Non-Human Identity, describes non-human identities as the identifiers assigned to code, applications, services and devices so they can authenticate and access resources without human intervention, and names lack of ownership, excessive permissions and limited visibility among their key risks.

Service accounts, API identities, workloads and AI agents raise the same questions: what the identity is, what it can reach, who owns it, why it needs that access, how its credentials are handled and when it should be reviewed or retired.

A machine identity without an owner is orphaned in all but name. For AI agents, a responsible human sponsor, a defined task and an expected lifespan should be part of the record from day one. Governing employees while ignoring machine and agent identities leaves a real hole in the control model.

Running the framework on Citadel Identity360

Citadel Identity360 governs people, machines and AI agents in one platform, whether the estate is cloud, SaaS, hybrid or on-premises, and each part of the framework has a matching capability.

Decision rights become routing. Requests, approvals, reviews and segregation-of-duties controls sit in one place, so the decision-rights map is applied by the workflow instead of by memory. Identities can also be grouped by vendor, so third-party access is governed supplier by supplier.

Lifecycle triggers become automation. Joiner, mover and leaver events drive provisioning and deprovisioning in connected and disconnected applications alike. For a system with no usable API, Citadel can use RPA to carry an approved change through the application’s own screens.

Ownership becomes visible. The identity graph links each person, account, role and permission, and access path discovery reveals how a given permission arrived and which approver signed it off. In cloud estates, Citadel’s CIEM capability collects entitlement scopes, permissions, grants, usage and policies, and presents an all-round view of the actions that need attention.

Risk sets the order of work. Risk scores can be tuned for the whole organisation, an individual identity, a department, an application or a manager, so tighter rules and earlier reviews land where the exposure is highest.

Reviews close the loop. In Citadel’s detailed review module, certification campaigns put a recommendation and risk context beside each access item, while identity attestation tests whether an identity should exist at all. Every revocation is followed until it is fulfilled.

Machines and agents get owners too. Each service account, API key and machine identity keeps an accountable owner from creation to retirement. Citadel pulls AI agents from the Microsoft Entra, AWS and Google Cloud registries, and its endpoint agents surface agents running on user devices, which helps uncover shadow AI. Every MCP server and agent is registered against an owner and authorised tool by tool; a configuration file generated in Citadel lets compatible agent platforms reach governed MCP services, and the kill switch lets a team step in and contain an agent. Non-human identity and AI-agent governance ship inside the product.

Deployment options cover SaaS, private cloud and on-premises, and each plan carries roughly 400 hours of customisation that can be spent shaping workflows around your own decision-rights map.

An illustrative week in a governed estate

Consider a finance analyst who moves into treasury on Monday. In Citadel Identity360, the HR change starts a mover workflow: treasury access is routed to the treasury application owner, and the old accounts-payable role is queued for removal. Until it goes, conflict detection flags the overlap between payment approval and vendor maintenance for the control owner. On Wednesday, a reconciliation agent the analyst built is registered with the analyst as owner and authorised for two named tools. When the quarterly campaign runs, the analyst’s manager sees risk context for each item, and each revocation is followed to completion. This analyst is an invented example, not a customer.

Measuring whether governance is working

To tell whether a framework is operating rather than sitting in a document, measure a few signals:

  • Time from an approved request to fulfilled access.
  • Time to remove or change access after a role change or departure.
  • Share of identities, applications and sensitive entitlements with a named owner.
  • Share of joiner, mover and leaver events handled through the defined path.
  • Orphaned or inactive accounts found, and how many were remediated.

These are not universal benchmarks. Track them against your own baseline. Our self-assessment guide shows how to set that baseline.

The takeaway for technology leaders

If access still depends on spreadsheets, email approvals and uneven follow-up, the gap is not only tooling; it is the absence of an agreed way to decide. Begin with visibility, policy and accountability, make lifecycle handling repeatable, then extend the same discipline to privileged access, machine identities and AI agents.

Identity governance framework questions answered

How does an identity governance framework differ from IAM?

IAM is the broader discipline of managing digital identities and their access, including authentication. An identity governance framework focuses on whether access is appropriate, approved, owned, reviewed and evidenced.

Is an identity governance framework something you can buy?

No. It is the organisation’s operating model: policy, ownership, process, controls and evidence. A platform such as Citadel Identity360 automates and records much of it, but the decisions about who decides what belong to the organisation.

Who should own the framework?

Overall ownership usually sits with the CIO, CISO or head of IAM, with named owners for policies, applications, entitlements and reviews beneath them. The key is that authority is written down and understood.

What review rhythm suits a governance framework?

No single interval suits every system. Set the rhythm by risk, sensitivity, policy and regulatory obligations, review privileged and high-risk access more often, and let tuned risk scores in Citadel show where to begin.

Is a bot or service account in scope as well?

Yes. Non-human identities need an owner, a justified purpose, lifecycle controls, credential handling and review or retirement. Citadel Identity360 handles them next to people, adding per-tool authorisation and containment for AI agents.

Put your own decision-rights map to the test

Take one application where nobody is sure who approves access, and list its owners, approvers and leavers from the last quarter. Bring it along and watch Citadel Identity360 turn that list into routed decisions, tracked removals and a clear record. Request Demo with that application in mind, or preview the platform in our product videos.

Stay Current

Get the latest insights delivered

Compliance updates, IGA best practices, and regulatory analysis from Astranova Labs.

Browse all posts →