Agents / Access Governance Agent
Meet your autonomous Access Governance Agent.
It discovers applications across your existing systems, establishes who is accountable for them, finds access that should no longer exist and coordinates its removal. Every outcome is verified where the application allows it, and every gap is reported.
You own the access decisions. Your agent does the work.
A closed ticket isn’t proof the access is gone.
Jordan Reyes · left Friday, 5:00 PM · deprovisioning ticket closed
HR (Workday)
Terminated, effective Friday
Terminated
Directory & SSO (Entra ID, Okta)
Account disabled, sessions revoked
Disabled
CRM
SCIM-provisioned, removed with SSO
Removed
Expense platform
Local admin account, outside SSO
Not checked
Plant scheduling tool
No API, no connector
Not checked
Access Governance Agent · re-reading each application
✓ Residual access removed and configuration verified · 1 application requires owner attestation
The person who pays for an application, the person who administers it and the person who should decide who gets access are often three different people, and one of them has often left. Watch your agent turn that into confirmed accountability, without sending thousands of people thousands of questions.
Application records from 7 sources
0Distinct applications
0Tier 1 applications
0Tier 1 with a confirmed owner
0Escalated to governance leads
0✓ 86% of Tier 1 owned and confirmed · inferred owners always labeled as inferred
Illustrative figures, not a customer environment.
The limit isn’t workflow automation. It’s the attention of thousands of people who don’t think access is their job.
While your IAM team sets policy and decides the hard cases, your agent keeps reconciling identities, owners and accounts across your estate, removes what policy says should go, and tells you plainly which applications it can and cannot verify.
What the agent detected
286 applications in SSO logs, expense and card data that no inventory knew about, 74 of them with local accounts that bypass SSO.What it did
Created a record for each with its users, how its access is managed and its probable owner, and labeled every application with its coverage state.What changed
Every application is classified. 1,940 without a connector are governed through scheduled extracts and owner attestation, not left out.What still needs a person
Approve the next onboarding wave, ranked by risk.What the agent detected
57 Tier 1 applications with no owner, an owner who left, or different owners in the IGA platform, the CMDB and the contract.What it did
Ranked candidates by evidence and sent each person one consolidated request. Followed every “not me” to the person named.What changed
34 reassigned, with owner records corrected in each system of record once confirmed.What still needs a person
23 applications with the Finance and Operations governance leads.What the agent detected
Nine new hires entered in the wrong business unit, about to receive that unit’s birthright access.What it did
Held the attribute-driven provisioning and told HR exactly what access each error would have granted.What changed
Seven corrected in HR and recalculated. HR stays the system of record; the agent never patches the directory around it.What still needs a person
Two records with HR.What the agent detected
Local accounts surviving central deactivation in SaaS consoles, for 38 people who left this quarter.What it did
Removed them through each application’s admin API and re-read every account. Raised tracked tasks where there is no API.What changed
1,106 removals configuration-verified in the target system. 17 applications reported as verification limited, each with an owner and a date.What still needs a person
Owner attestation for the 17.What the agent detected
3,480 accounts not tied to any person, four of them used after their owner’s termination date.What it did
Linked high-confidence matches, sent the rest to application owners, and preserved the activity evidence for the four before touching them.What changed
2,910 linked, reclassified as service accounts with an owner, or removed.What still needs a person
Security is reviewing the four as potential incidents.What the agent detected
Six accepted exceptions reaching expiry, two of them renewed twice before.What it did
Asked each owner to remediate or renew with fresh justification, and checked each approver’s authority for that risk level.What changed
Every action carries its source, time, decision-maker and verification result, retrievable by control, application and period.What still needs a person
Risk acceptance. The agent never accepts risk on your behalf.Effective access verifiedcan no longer be usedConfig. verifiedread back from the target systemVerification limitedreported as an exception, with an owner
Illustrative data, not a customer environmentAutonomous IT, finally. · xops.io
Role
Identity and access governance operations
Works across
Workday, Entra ID, Okta, your IGA and PAM platforms, ServiceNow, SaaS admin consoles, Teams and Slack
Authority
What you delegate, within your access policy
Measured on
Inappropriate access removed and verified, ownership coverage, remediation speed, audit outcomes and human effort saved
Responsibility
What it does
Done means
1One trusted identity
“Why do HR and the directory disagree?”
What it does
Reconciles HR, contractor and directory records into one identity per person. Reports discrepancies to the owner of the source record and holds access that depends on the disputed data.
Done means
Every identity is linked to an authoritative record, or scheduled for disablement.
2Application coverage
“Which SOX applications aren’t governed?”
What it does
Discovers the applications that grant access, from SSO, expense, card spend and your software inventory, and records how far each one is under governance.
Done means
Every discovered application has a tier, a coverage state and an onboarding decision.
3Ownership
“Who should approve access to this?”
What it does
Establishes the business owner, budget owner, technical owner and access approver for each application from evidence, and gets each one confirmed by the person.
Done means
A confirmed owner, recorded the same way in every system of record.
4Leaver and mover assurance
“Does anyone who left still have access?”
What it does
Finds access that survived central deactivation, including local application accounts, removes it through the available path and re-reads the target system.
Done means
Each removal is configuration-verified in the target system, effective access is verified where the evidence allows, and anything else is an open exception with its reason and owner.
5Orphaned and dormant accounts
“Who does this account belong to?”
What it does
Ties every readable account to an accountable identity. An account used after its owner left is preserved as evidence and escalated to security, not just disabled.
Done means
Every account is linked, reclassified with an owner, or removed.
6Reviews people can actually decide
“Why am I reviewing 1,800 items?”
What it does
Prepares each review with what the access does, when it was last used and a recommendation, and verifies every revocation once the campaign closes.
Done means
Fewer, better decisions, and every revocation confirmed in the target system.
7Privileged and non-human access
“Which service accounts have no owner?”
What it does
Finds privileged access held by leavers and movers, and gives service accounts, keys and AI agents an accountable owner. Orchestrates through your PAM platform rather than replacing it.
Done means
Every privileged and non-human identity has a confirmed owner. Removals follow approved disablement, dependency and recovery controls.
8Exceptions and audit evidence
“Can we prove every leaver lost access?”
What it does
Turns every violation into a remediation, a time-bound exception or an escalation, and records the evidence for every control as it runs.
Done means
Evidence retrievable by control, application and period, released only with the control owner’s approval.
Your identity platforms run the workflows. Your agent makes sure the outcomes happen.
Your directory and identity provider remain the system of record for identities and authentication. Your IGA platform keeps running the certifications and approvals it runs today, and your PAM vault keeps the privileged credentials. Your Access Governance Agent works across them and fills the gaps between them: the applications they don’t reach, the owners nobody confirmed and the removals nobody checked.
Your team retains control. Access policy, approval thresholds, separation-of-duties rules, risk acceptance and exceptions remain with authorized people. The agent never completes an attestation itself or assigns an owner without that person’s confirmation.
Every application carries an explicit coverage state, and every report is qualified by it. Where the agent can’t read or verify an application, it says so, names an owner and manages the exception. It never answers “is everyone who left gone?” with an unqualified yes.
1
Discovered
Knows the application exists, and from which sources.
2
Identity-correlated
Can tie its accounts to people.
3
Entitlement-visible
Can read what each account is permitted to do.
4
Writable
Has an authorized path to change access directly.
5
Independently verifiable
Can confirm a change from the application, not the system that made it.
Or
Manually governed
Owner attestation or a tracked manual task, reported as such.
Requested
A ticket was raised or an API call was made.
Not done.
Executed
The system that made the change says it worked.
Still not done.
Configuration verified
The target system itself confirms the requested change, with the evidence, its source and its time recorded.
Done, as configured.
Effective access verified
The available evidence confirms the access path can no longer be used, including sessions, tokens and alternate routes.
Done.
Verification limited: where the agent can’t establish the result, it records the remaining exposure, names an owner and reports it as an exception.
Your access policy decides what the agent does on its own, what it asks first and what it escalates. People set the policy and own the exceptions.
Does on its own, within your policy
Removing a leaver’s access at the specified time, disabling accounts tied to terminated identities, and reclaiming unused access when approved policy, verified inactivity and the safety checks permit.
Asks before it acts
Assigning an owner, removing privileged access, deleting a service account, releasing evidence to an auditor.
Escalates to your team
An account used after its owner left, an expired exception the owner won’t resolve, an acquisition, a control failure in an in-scope application.
It never accepts risk on your behalf, completes an attestation, approves its own proposals or changes its own permissions.
Before any removal
Authority to remove access isn’t enough. Every removal passes these checks first.
If any check can’t be completed, the action drops to asking first.
✓ Cleared, removed, configuration verified
Your CISO should ask. The agent is itself a governed identity, with a named owner in your IAM organization, and its authority is set by your people, not by XOPS and not by the agent.
The model proposes. Policy authorizes.
A model output alone can never change access. Every action passes deterministic checks on trigger, authority, safety and approval.
Starts read-only. No standing admin.
A separately scoped identity for each connected system, with read and write kept apart and credentials never placed in prompts, tickets or logs.
Approvals bound to one action.
The executed change must match what was approved exactly. Any change in scope voids the approval.
One switch to stop it.
Suspend write actions for the whole deployment, one system or one kind of action, while monitoring carries on.
Content is data, not instructions.
Text in a ticket, email or entitlement description that tries to direct the agent is flagged to security and changes nothing.
Every step on the record.
Proposals, approvals, credential use, actions and verification in a tamper-evident log you can export to your SIEM.
When someone leaves, the Workforce Lifecycle Manager owns the transition. The Access Governance Agent supplies the access model and the verification, and the agents that already know your software, devices and AI agents each close their part.
Who is accountable for what
Access Governance Agent
Identity correlation, application and entitlement ownership, reconciliation, orphaned accounts, access policy violations and access audit evidence.
Workforce Lifecycle Manager
Joiner, mover, leave and leaver outcomes across identity, devices, software, communications and workplace.
Software Asset Manager
Licenses, subscriptions and software ownership for commercial purposes. Removed access frees licenses.
AI & Agent Governance
AI agents’ model access, execution policy, budgets and cost. The access agents hold is governed here.
Hardware Asset Manager
Physical devices. Device-based access is coordinated with the Access Governance Agent.
Your agent doesn’t need to govern thousands of applications on day one. Start with a defined set of high-priority applications and the systems you already trust. It discovers conflicting ownership records, identifies likely owners, coordinates confirmations and measures how much manual effort it removed. Expand coverage as the results are proven.
01 · Establish
Discover and reconcile
Build the application and ownership baseline from the systems you already trust, with every conflicting record surfaced.
02 · Confirm
Engage the right people
Resolve ambiguous ownership with one consolidated request per person, following referrals and escalating only what remains.
03 · Operate
Keep it current
When an owner changes role or leaves, a contract changes hands or a new application appears, the agent detects the drift, asks the right person and verifies the record.
Measured from the start: ownership coverage, conflicting records resolved, confirmations people had to give, elapsed time and drift caught after go-live.
Applications appear. Owners change roles. Accounts outlive the people who held them. Reviews get rubber-stamped.
Your Access Governance Agent establishes who owns every important application, finds the access that should no longer exist and brings your team only the decisions that need a person.
Every approved action runs through the identity platforms you already have. Every outcome is verified, or reported honestly when it can’t be.
