Skip to content

Managing Playbooks

Shadow AI Remediation provides customizable GenAI playbooks that enable security teams to automate responses to various GenAI user actions and engage users at critical decision points to improve security outcomes.

Playbooks run as just-in-time workflows that appear as a messenger in the users’ session, guiding them toward compliant behavior when interacting with GenAI tools.

Remediation playbooks enable:

  • Prevention of risky GenAI usage.
  • Guidance for users toward sanctioned tools.
  • Reinforcement of corporate GenAI policy.
  • Capturing of user intent when policy violations occur.

Shadow AI Remediation includes four out-of-the-box playbook templates:

Important

Shadow AI Remediation is designed to help guide employees toward AI policy compliance that is in line with your organization’s overall policy efforts on AI. It is not designed to catch all instances of bad behavior, intentional or willful violations, or malicious activity within your organization. This product does not replace the need for your organization to have comprehensive auditing, human oversight, or robust, industry-standard security measures.

Prerequisites

To ensure GenAI applications are permitted and blocked inline with your organizations AI policy, you must determine which playbooks should be enabled and complete any supporting configuration.

To configure Shadow AI Remediation to meet your organization’s AI policy:

  1. Define the list of sanctioned applications

    • SaaS
    • Home-grown

  2. Prepare relevant URLs for the playbooks:

    • Note the URL of your sanctioned tool(s).
    • Note the URL of your organization's AI policy.

  3. Review policy-related considerations

    • Are you blocking commonly-banned GenAI tools such as DeepSeek?
    • When using other non-sanctioned AI tools, how firm would you like to be in redirecting the users to the sanctioned tool?
    • If you permit users to override the policy and use a non-sanctioned tool:
      • How often would you like to pop up and alert when users use the non-sanctioned tools?
      • Would you like to further alert when a file upload begins?

Defining Sanctioned Applications

If the GenAI tool is a well-known mapped SaaS application, such as Microsoft Copilot, the application must be added to the generative-ai-sanctioned-apps list.

If the GenAI tool is not a well-known mapped SaaS application, create a Custom Application to add the application, and then add it to the following lists:

  • generative-ai-apps - Ensures the application will be included in the list of sanctioned GenAI tools.
  • generative-ai-sanctioned-apps - Ensures playbooks will not interfere when users are using the application.

If you are unable to add a custom application because its URL overlaps with an existing domain of a SaaS application in the global application database, add the FQDN (hostname) of the application to the following list instead:

  • generative-ai-sanctioned-hostnames - Ensures playbooks are not triggered when visiting the home-grown applications.

Tip

A correct FQDN example includes:

  • copilot.microsoft.com

Incorrect FQDN examples include:

  • https://copilot.microsoft.com/ - the HTTPS scheme is not part of the FQDN.
  • copilot.microsoft.com/prompt - the prompt is part of the path, not a valid FQDN.

Use the below considerations to determine the playbooks to use to match your organization’s policy.

Consideration Recommended Action
Are you blocking commonly-banned GenAI tools such as DeepSeek? Configure and enable the Avoid Banned GenAI Applications playbook.
When using non-sanctioned GenAI tools that are not explicitly banned, should the user be blocked from using the application? Block
Configure and enable the Avoid Non-Sanctioned GenAI Applications playbook.

Permit with justification
Configure and enable the Corporate GenAI Policy Guidance playbook.
For users permitted to override the policy and use a non-sanctioned application:
How often should the messenger and alert be displayed to users when using non-sanctioned tools? Configure Throttling to limit the number of times the playbook should be triggered for users.
Display an additional alert when a file upload begins? Configure and enable the File Upload to Non-Sanctioned GenAI playbook.

Deployment Best Practice

SailPoint recommends a staged rolled out approach to ensure the configured playbook runs as expected, and prevent unintended blocking of sanctioned AI tools.

To prevent unintentional blocking of user tasks, SailPoint recommends the following rollout approach:

  1. Assign the playbook to a small target group first.

  2. Validate the playbook experience and messaging.

  3. Confirm sanctioned tool redirection works as expected.

  4. Expand deployment to broader user base.

Important

To prevent unintentional blocking of user tasks, SailPoint strongly recommends a staged rollout approach.

Playbook Catalog

The playbook catalog includes out-of-the-box playbook templates for enforcing corporate policy. Playbook templates must be configured to meet your organization’s policy, culture, and set of sanctioned applications. Playbook templates are added to Company Policies where they can be configured to meet your organization’s needs.

Note

When adding a playbook to the company policies, you must select a target group to control which users the playbook will apply to. SailPoint strongly recommends initially assigning a test target group containing limited users while the playbook is tested. Ensure you have created a test target group before adding the playbook to the company policies.

To add a playbook template to the company policies:

  1. Go to Playbooks > Playbook Catalog.

  2. Select the desired playbook and select Add.

  3. (Optional) To restrict the existing match criteria to a specific application, within the Conditions section, select And to add a new And condition.

    Warning

    Do not remove the default match conditions. Removal of the default match criteria may result in unintended blocking of applications causing significant productivity disruptions.

    • Within the Conditions section, complete the following:
      • Select the first field and choose Application.
      • Leave the second field as In.
      • Select the Value field and select Application.
      • Select the desired application(s), and select Add new app.

  4. (Optional) Within the Trigger match criteria window, select Enable throttling to configure the number of times within a defined time period the playbook should trigger.

  5. Select Next.

  6. Within the Edit target group window, select the Target group field.

  7. Select the desired target group.

    Tip

    Select Show next to a target group to view its members.

  8. Select Next.

The playbook is available to edit and enable in the company policies page.

Reviewing User Requests

If permitted, users can request for non-sanctioned GenAI applications to be sanctioned.

To review user requests:

  1. Go to Logs > GenAI.

  2. Select the Database icon  next to the Log search box and select Saved queries.

  3. Within the Saved queries window, select User justification for non-sanctioned tools.

All requests will be displayed. Selecting a log entry opens the detail panel with detailed contextual information including the justification provided by the user within the User Response field.

Throttling Playbook Triggers

If you do not want a playbook to trigger every time its match criteria is met, you can configure the number of times within a defined time period the playbook should trigger.

To throttle the number of times a playbook is triggered:

  1. Go to Playbooks > Playbook Policies.

  2. Select Actions > Edit match criteria.

  3. Within the Trigger match criteria window, enable Throttling.

  4. In the time(s) field, enter the number of occurrences the playbook should trigger for a user.

  5. (Optional) Select + Add time period to define the number of occurrences the playbook should trigger for a user within a desired time period.

  6. Select Save.