Table of Contents

1. Section 1: Environment Setup & Prerequisites
2. Section 2: Requesting AI Use Cases
3. Section 3: Troubleshooting AI Governance & Assessments
4. Section 4: The AI Use Case Lifecycle
5. Section 5: Backend Architecture & Tables

Section 1: Environment Setup & Prerequisites

The AI Control Tower is a governance module in ServiceNow designed to regulate Artificial Intelligence applications across an instance. It provides visibility into backend operations by tracking shadow AI usage, monitoring table and data access, and enforcing governance policies.

Step-by-Step Environment Setup

Provision a Personal Developer Instance (PDI) on the ServiceNow Developer portal. For optimal compatibility, select the Australia release version. During plugin installation, ensure the “Load demo data” checkbox is selected for the following essential modules:

• AI Control Tower
• AI Control Tower Core
• AI Risk and Asset Management for ServiceNow AI
• Smart Assessment Core
• Smart Assessment Collaboration
• Advanced Response Automation for Smart assessments
• Basic Scoring for Smart Assessments
• AI Risk and Compliance Management
• AI Case Management
• Content Publishing
• AI Trace Collector

Required Permissions

Managing AI use cases requires specific governance privileges. The primary required role is AI Steward (sn_ai_governance.ai_steward). System administrators must explicitly assign this role to their accounts or designated test accounts to ensure full management capabilities.

Section 2: Requesting AI Use Cases

Submitting a Request

To begin onboarding AI assets into the governance workflow, execute the following procedure:

• Navigate to the Service Portal and open the Request an AI Use Case record producer.
• Fill in the AI use case name (for example, “go to the moon”) and define the version number (e.g., 1.0.0).
• Select the appropriate model category from the available options, such as Agentic AI.
• Specify the Provider, which acts as the manufacturer of the AI, such as Google.
• Assign the Managed by field to a specific user. The person selected here will be responsible for managing the AI asset moving forward, handling tasks like future recertifications.
• Once submitted, you can click on the confirmation to open the newly created record in the backend.

Transitioning from Unmanaged State

When a new AI use case is submitted, it is automatically tagged as Unmanaged and will not immediately appear in your main dashboards.

The AI Steward is responsible for reviewing the description and details of the unmanaged asset. By clicking Start Review, the lifecycle state transitions to In Review. From this point forward, the designated “Managed by” user can initiate the necessary assessments and continue the workflow.

Section 3: Troubleshooting AI Governance & Assessments

A common issue during asset submission is that the impact assessment fails to trigger or does not display the expected Take Assessment or Start action buttons. Resolving this requires verifying background flows, templates, and permissions.

Diagnosing Flow Designer Executions

When an AI use case enters the Assess phase, an out-of-the-box flow named AIRC flow once AI Gov Details is in assess phase (or related governance flow) is responsible for:

• Creating or updating the AI system record.
• Evaluating whether existing records match the submission.
• Triggering the Fundamental Rights/Impact Assessment for AI Assessments.

To check where things are failing:

1. Navigate to Flow Designer > Executions.
2. Locate the execution corresponding to your submitted record.
3. Turn on flow reporting if you haven’t already (via Flow Administration > Properties) so you can trace individual steps.
4. Verify that the flow ran through to completion and that the Create Assessment action executed without runtime errors.

Checking Template and Entity Configurations

Even if the flow runs cleanly, the assessment interface may remain blank if the supporting configuration records are not published or active.

• Assessment Workspace Templates: Navigate to Workspaces > Assessment Workspace. Locate the template (such as the Fundamental Rights Impact Assessment for AI). Verify that the template status is explicitly set to Published.
• Assessment Purpose & Scope: Open the assessment purpose record linked to the AI system. If you are developing inside an application scope, ensure your application picker matches the required scope so changes are saved properly.
• Entity Status: Check GRC > Entities to verify that target entities associated with your AI use cases are set to Active.
• Risk Methodologies: Confirm that methodologies like the Risk Assessment for AI Inventory and Risk Classification for AI are fully configured and published.

Resolving Plugin Installation Failures

Installing AI Control Tower plugins can be time-intensive. If a plugin (such as Smart Assessment) fails midway, navigate to the plugin record and select Repair to force a clean reinstallation.

Section 4: The AI Use Case Lifecycle

1. Initiating the Review Process

• Open the unmanaged AI asset record.
• Click the Start Review button.
• The system will automatically transition the record’s lifecycle state to In Review.

This transition triggers the backend workflow to generate the necessary compliance tasks, assigning them to the relevant stakeholders.

2. Completing the Impact Assessment

Once the record is in review, the primary task is to complete the generated assessments.

• Navigate to the assigned tasks and click Take Assessment.
• Depending on your governance setup, this could be the Fundamental Rights Impact Assessment or the EU AI Act (EUAI) compliance assessment.
• Fill in the required questionnaire. You will typically need to answer conditional Yes/No questions, such as whether the AI system processes sensitive personal data or requires specific risk controls.
• Submit the answers and explicitly mark the assessment task as Complete.

3. The Build and Test Phase

With the initial review and impact assessment out of the way, the AI Use Case automatically advances to the Build and Test state.

During this phase, the focus shifts from theoretical risk to practical implementation:

• Development Plans: Teams must outline how the AI model will be safely constructed and tested.
• Pre-Deployment Validations: Additional assessments and control attestations are generated to ensure the build aligns with the approved parameters.
• Background Orchestration: ServiceNow’s Flow Designer (specifically the AIRC Governance flows) continues to run in the background, validating that all conditions are met before allowing the asset to move forward.

4. Final Approval and Deployment

After all testing requirements and pre-deployment assessments are satisfied, the AI Use Case enters the final stretch.

• The system routes the record for Approval to the designated risk and compliance managers.
• Once all required sign-offs are captured, the lifecycle state officially updates to Deployed.
• The AI system is now fully active, governed, and continuously monitored by the AI Control Tower.

This structured workflow ensures that no AI model slips into production without rigorous documentation, testing, and approval.

Section 5: Backend Architecture & Tables

Understanding the core backend architecture is essential for customizing the AI Control Tower, debugging automated workflows, and configuring Smart Assessments. Below are the core tables governing AI asset management and assessment execution:

• AI System Digital Asset (alm_ai_system_digital_asset): This is the primary table where your new AI use case record is stored.
• Governance (Sn_ai_governance_asset_governance_details): This critical table tracks the heartbeat of your use case. It records the lifecycle phase, status updates, phases, and state transitions (like moving from Unmanaged to Deployed).
• AI Governance Task (Sn_ai_governance_assessment_task): This handles the approval tasks and acts as the binding glue for impact assessments and AI Asset.
• Assessment Template (Sn_smart_asmt_template): All assessment templates will be stored here
• Assessment Question Instance (sn_smart_asmt_question_instance): All user assessment responses will be stored here.
• Assessment instance to persona to users (sn_smart_asmt_m2m_instance_persona): Stores details related to Assessment Contributor details.
• Assessment instance to scope item (Sn_smart_asmt_m2m_instance_scope_item): Connects the Assessment to the Task/AI Asset
• Response automation predicate (sn_smart_asmt_response_automation_predicate): Stores logics of Autopopulate/Automated Responses of Question
• Response option (sn_smart_asmt_response_option): Stores options for questions defined in an assessment template.
• Section (sn_smart_asmt_section): Stores section and its templates.
• Question (sn_smart_asmt_question): Stores the questions in an assessment template.

Process Orchestration with Playbooks

Behind the scenes, the AI Asset Onboarding process relies heavily on ServiceNow’s Playbook experience (Process Automation) and Flow Designer to seamlessly route records through these tables and generate the appropriate tasks. By mastering this backend structure, you can move beyond the out-of-the-box configurations and tailor the AI Control Tower to your exact requirements.