Skip to content
English
  • There are no suggestions because the search field is empty.

Approve or reject an auto-remediation

Summary

Where auto-remediation is enabled, the platform can propose automated playbook steps on an incident that need a human sign-off before they run. This step completes when you review one proposed automatic fix and approve or reject it.

Before you start

Auto-remediation depends on your plan, and only appears once your organization has it enabled.

You also need the right permissions:

  • Approve and continue and Re-run playbook need the Manage Endpoint Protection permission, because a playbook step can block a person or an AI tool for your whole organization. If a step would withdraw an AI tool's approval for your whole organization, approving it also needs Manage Platforms.
  • Reject and stop needs Manage Endpoint Protection too, because rejecting a step is the other half of the same decision.

Two further conditions decide whether the panel is on the incident in front of you. The incident has to be confirmed real, which is step 1 below. It also has to carry proposed playbook steps: on an incident where no steps were produced, the panel is absent rather than empty, and there is nothing to approve.

1. Confirm the incident is real

  1. Go to Inbox → Incidents and expand the row of an incident that has automated playbook steps proposed. The approval panel is in the expanded row, not on the incident's own page.
  2. An open incident leads with Is this a real incident?. Choose Confirm real.

False positive and Defer both close the incident out without running the playbook, and the approval panel stays hidden on those paths.

2. Find a proposal

The Auto-remediation panel lists each automated step awaiting sign-off. Where every proposed step is a manual one, the same panel is headed Recommended actions.

3. Decide

For each proposed step:

  • Approve and continue: runs the step and moves to the next.
  • Reject and stop: halts that step and the playbook run.

Once every step has finished, Re-run playbook on the same panel runs the whole sequence again.

Confirm it worked

The incident's automated timeline shows the step you approved or rejected, and its status reflects the outcome: executed, or stopped.

If it didn't work

  • No auto-remediation panel appears: the panel stays hidden until you answer Is this a real incident?. Check the incident shows Disposition: Confirm real near the top, and use Change if it is set to something else. If the incident is confirmed real and the panel is still absent, this feature depends on your plan and on auto-remediation being enabled for your organization: check both with your account team.
  • Told to refresh the incident and approve (or reject) the step shown: the playbook moved on since you opened it, so you were looking at an older step. Refresh, then decide on the step now shown.
  • The panel isn't on the incident page you opened: the approval panel is in the expanded incident row under Inbox → Incidents, not on an incident's own page.
  • The incident is confirmed real and there is still no panel: this incident carries no proposed playbook steps. The panel appears only where steps were produced for that incident, so open one that shows a named playbook with steps listed.
  • The panel is headed Recommended actions and offers Run this playbook: the steps on this incident are recommendations that run only when you start them. There is nothing to approve or reject: use Run this playbook to run the sequence.
  • Approve is refused: you need the Manage Endpoint Protection permission, and Manage Platforms too if the step withdraws an AI tool's approval for your whole organization. Ask an administrator.
  • Rejected a step and the whole playbook stopped: this is expected: rejecting halts the run rather than skipping just that one step. Re-run the playbook from the panel if you want to continue past it manually.
  • Unsure whether to approve: the incident's enrichment panels (User Context, Baseline & Anomalies, Evidence) provide the same context used elsewhere in the incident; review those before deciding if the proposed action isn't self-explanatory.

Related