# Access review remediation: a worked exception | SourceX

Canonical: https://refer.sourcex.si/guides/customer-access-review-remediation-exception-examples
Last updated: 2026-11-04

The records that explain a customer access review exception are the ones that show, in order, the problem, the authorized action and the checked result: the review item that flagged it, the finding and the owner's written response, the approved change ticket, the directory audit event proving the removal, and the next cycle's re-test. A list of current permissions shows only a state. The case below walks through one exception, a service account still owned by someone who has left, step by step.

This walkthrough is part of the [customer access review remediation guide](/guides/workflow-customer-access-review-remediation), which covers what to inventory before an identity tenant merger. Everything here is **Illustrative**: a fictional company, invented keys and dates.

## The case: an orphaned service account with customer data scope (Illustrative)

A 640-person payments software company runs quarterly reviews of who can reach customer tenants and customer data. Reviews run in its identity provider and are assigned to resource owners, with managers as a fallback. Microsoft Entra, for example, lets a review go to resource owners, delegates, managers or users themselves, with fallback reviewers when nobody else is available.

In the January cycle, one item stands out. A service account that runs nightly reconciliation exports still lists its owner as a finance-systems manager who left two months earlier. The owner's account is disabled. The service account is active and can read customer data. Nobody is left to vouch for it.

This is a hard case because the obvious fix, disabling the account, would break a reconciliation job that finance depends on. Someone has to decide, not just click.

## Step by step: the decision, the decider and the record

| Step | Date | What happened | Who decided | Record | System |
|---|---|---|---|---|---|
| 1. Finding | Jan 14 | Review item flags owner disabled, account active, customer data scope | Fallback reviewer | Review decision with comment | Identity provider review |
| 2. Escalation | Jan 16 | Finding SEC-212 opened and assigned | GRC owner | Finding record | GRC tool |
| 3. Owner response | Jan 21 | Account still feeds reconciliation; propose new owner and read-only scope | Data platform lead | Written response on SEC-212 | GRC tool |
| 4. Authorized action | Jan 28 | Change CHG-4471 approved: reassign owner, rotate credential, remove write scope | Security operations head | Approved change ticket | Ticketing system |
| 5. Removal | Jan 29 | Owner changed; write role removed | Identity administrator | Audit events | Identity provider audit log |
| 6. Re-test | Apr 9 | April review: new owner approves with justification | New owner | Review decision with justification | Identity provider review |

Read the table from the bottom up and each row answers a question an auditor, a customer or a model evaluator would ask. Did the fix hold? Step 6. Did it really happen? Step 5. Who allowed it? Step 4. Why not just disable it? Step 3.

## Why step 5 is the record most likely to vanish

Steps 2 to 4 sit in the GRC tool and ticketing system, which usually keep years. Step 5 sits in the identity provider's audit log, which usually does not. Microsoft Entra keeps audit logs for 30 days on P1 and P2, and 7 on Free, and expired data cannot be recovered unless it was archived. Microsoft documents archiving those logs to a storage account through diagnostic settings for exactly this reason.

So by the April re-test, the January removal event may already be gone from the identity provider itself. It survives only if the company archived those logs before the window closed. Without it, the company can say it fixed the problem but cannot show it. That one question, "do you keep identity audit events beyond the default?", tells you a lot about whether a company's remediation chains are complete.

## What the full chain teaches, and what weakens it

A complete chain has the shape that makes operational records useful for AI training and evaluation: a problem, a reasoned decision by someone with authority, an action and a verified outcome. An AI agent asked to handle access exceptions would need to learn exactly when to escalate, when to keep access with narrower scope, and how to document it. [Why AI agents need work data](/proof/ai-agents-need-work-data) explains the general case.

| Weak link | What is lost |
|---|---|
| Finding closed "fixed" with no ticket key | Link from decision to action |
| Review decision with no justification | Why access was kept or removed |
| Owner response by chat, not on the finding | The business reason |
| Audit events expired before export | Proof the action happened |
| Re-test in a new tenant with no mapping | Proof the fix held |

The last row is common after tenant mergers, when accounts are rebuilt and lose their history. The [records across systems map](/guides/customer-access-review-remediation-records-across-systems) shows the identifiers that keep a chain joined.

## How to recognize similar cases at a client

You do not need to see a finding to know whether a company has chains like this. Ask the security operations head, in conversation:

1. Do access reviews cover customer-facing systems or customer data, and how often?
2. Are service accounts included, with named owners?
3. Do findings live in a GRC tool, and do they cite change ticket keys?
4. Are reviewers required to justify decisions?
5. Are identity audit logs archived beyond the provider's default, and since when?
6. Has a tenant merger or identity migration happened or been planned?

Yes to the first five suggests years of complete chains. A yes on the sixth makes the question of what survives concrete. This differs from auditor-driven remediation on the [management letter comments page](/data-types/management-letter-comments-remediation): access reviews are the company's own control from the first record. If the company has had a security incident, [whether it can still license data](/questions/can-a-company-with-a-data-breach-license-data) is a separate question. Similar exception patterns in operations appear in [logistics exceptions and their documented resolutions](/guides/how-to-identify-logistics-exceptions-and-their-documented-resolutions).

## Qualification and what stays unknown

Whether the records can be licensed depends on rights, qualification and the proposed use. Review records name employees and customers, and customer contracts may limit reuse, so the rights review comes first. The [who-can-license page](/questions/who-can-license-customer-access-review-remediation-records) covers which executive to introduce.

The introduced company needs 50+ full-time employees at peak (contractors excluded), a documented operating history, rights to license and an authorized contact. Companies in the U.S., Canada and the EU qualify, as do companies in the UK, Australia, New Zealand, India, Singapore, South Africa, Latin America, the UAE and Saudi Arabia when at least 70% of the data is in English. Partners sign the Channel Partner Agreement before submitting introductions, and the first valid introduction within the 90-day window is credited when the company submits a verified application. Whether a buyer wants a given company's chains is known only after SourceX reviews them.

## Next step

> **Refer a company. No cap on rewards.** Default reward on a company's first qualifying transaction: the greater of $50,000 plus 5% and 15% of Actual Supplier Payout, with no cap; further qualifying transactions first agreed within 12 months of the first Closing pay 15%. Eligibility and payment conditions apply, including the USD 100,000 Minimum Transaction, and a signed Schedule A or addendum can differ. A lead, meeting or signature alone does not trigger payment.

If a client's findings cite tickets and its identity audit events are kept beyond the default, [register as a partner](/register), check the company against [who qualifies](/who-qualifies) and read the [program terms](/terms) before you introduce the security operations head.

## Sources

- https://learn.microsoft.com/en-us/entra/id-governance/access-reviews-overview
- https://learn.microsoft.com/en-us/entra/identity/monitoring-health/reference-reports-data-retention
- https://refer.sourcex.si/terms
- https://learn.microsoft.com/en-us/entra/identity/monitoring-health/howto-archive-logs-to-storage-account
