AI Guardian is ServiceNow's governance platform for organizations using artificial intelligence features like Now Assist or custom AI models on the platform. It addresses a growing concern: when AI systems generate responses automatically for employees or customers, how do you ensure those responses are appropriate, unbiased, and compliant with company policies? AI Guardian monitors what AI systems are saying, flags problematic content, and gives administrators control over AI behavior across ServiceNow. Once AI Guardian is active, it sits between your AI models and end users, checking every AI-generated response against your organization's rules. This means you can deploy AI features like automated incident summaries or chatbot responses while maintaining oversight of what gets communicated. The system creates an audit trail of all AI interactions and provides dashboards showing where AI might be creating risks or operating outside expected parameters.
Key Capabilities
Output Content Filtering
Automatically blocks or flags AI responses that contain inappropriate content, sensitive information, or violations of company policies. You set the rules, and AI Guardian enforces them before responses reach users. This prevents AI from accidentally sharing confidential data or generating problematic content.
AI Interaction Monitoring
Tracks every conversation and response generated by AI systems across ServiceNow, creating a searchable record of what AI told users and when. Administrators can review patterns, identify recurring issues, and understand how AI is being used across different departments.
Bias Detection and Alerts
Analyzes AI responses for potential bias in language, recommendations, or decision-making patterns. When the system detects responses that might treat different groups unfairly, it alerts administrators and can automatically flag the interaction for human review.
Risk Policy Management
Lets you define specific rules about what AI can and cannot say or do in different contexts. For example, you might allow AI to suggest solutions for IT incidents but require human approval before AI responds to HR inquiries. Policies can be customized by department, user group, or type of request.
Compliance Reporting
Generates reports showing how AI systems are performing against regulatory requirements and internal standards. This includes tracking what data AI accessed, how it was used, and whether all interactions followed established protocols. Essential for audits and regulatory compliance.
Model Performance Analytics
Provides dashboards showing how well AI models are working, including response accuracy, user satisfaction scores, and frequency of human interventions. This data helps identify when models need retraining or when policies need adjustment.
How It Works
When a user interacts with an AI feature in ServiceNow, such as asking Now Assist to summarize an incident or requesting help from a chatbot, AI Guardian intercepts the AI's response before it reaches the user. The system checks the response against your configured policies, scans for sensitive information or inappropriate content, and analyzes it for potential bias. If everything looks acceptable, the response goes through to the user immediately. If AI Guardian detects issues, it can block the response, send it for human review, or log it as a policy violation while providing a generic fallback message to the user.
Who Uses It and How
Healthcare system
Uses AI Guardian to ensure their IT helpdesk chatbot never accidentally shares patient information or medical details when helping staff with technical issues. The system monitors all chatbot conversations and blocks any responses containing health data.
Result: Maintains HIPAA compliance while allowing AI to handle routine IT requests automatically.
Financial services company
Deployed AI Guardian to monitor their Now Assist implementation, which helps analysts summarize security incidents. The system ensures AI summaries don't contain customer account numbers or trading information that could create compliance issues.
Result: Reduced incident resolution time by 40% while maintaining regulatory compliance across all AI-generated content.
Global manufacturing company
Uses AI Guardian to track bias in their AI-powered employee request system, ensuring that AI responses about vacation requests, equipment access, and facility services treat all employees fairly regardless of location or role.
Result: Identified and corrected AI bias that was prioritizing requests from certain office locations over others.
University IT department
Implemented AI Guardian to monitor their student-facing AI chatbot that helps with password resets and account issues. The system prevents the chatbot from sharing sensitive student information and ensures responses are appropriate for an educational environment.
Result: Enabled 24/7 automated student support while maintaining FERPA compliance and student privacy protection.
Sourdough: ServiceNow Monitoring and Analytics
A Chrome extension for ServiceNow Admins and Developers with essential tools, analytics, graphs and monitoring features.
Free to install. Pro $5/month after a 14-day no-card trial.
Pro requires the ServiceNow admin role. Upgrade inside the extension.
Implementation: What to Know
Implementation typically requires coordination between IT security, compliance, and ServiceNow administrators, with initial setup taking 2-4 weeks for most organizations. You need existing AI features deployed in ServiceNow for AI Guardian to monitor, so this usually comes after Now Assist or custom AI models are already in use. The most common implementation delays happen when organizations try to create overly complex policies from the start rather than beginning with basic content filtering and expanding gradually. Start with simple rules about blocking sensitive data types, then add more sophisticated bias detection and custom policies as your team learns what patterns matter most for your organization.
Common Use Cases
Preventing AI from sharing sensitive customer data
A customer service AI chatbot helps agents draft responses to customer inquiries. AI Guardian ensures these draft responses never include customer social security numbers, account balances, or other sensitive information that might accidentally be included in AI training data.
Monitoring AI bias in employee services
An HR AI system helps employees find information about benefits and policies. AI Guardian tracks whether the AI provides consistently helpful responses to all employees or if it gives better service to certain groups, departments, or geographic locations.
Ensuring compliance in regulated communications
A financial services company uses AI to help draft incident response communications. AI Guardian reviews each AI-generated message to ensure it follows regulatory language requirements and doesn't make claims that could create legal liability.
Auditing AI decision-making in IT operations
IT teams use AI to automatically categorize and prioritize incoming tickets. AI Guardian creates a record of every AI decision, including what data the AI used and why it made specific recommendations, supporting accountability and continuous improvement.
Controlling AI access to confidential information
A company's AI system helps employees find documents and information across ServiceNow. AI Guardian ensures the AI only references and shares information that users are authorized to see, preventing accidental disclosure of confidential data.
Key Tables
sys_ai_guardian_policysys_ai_interaction_logsys_ai_audit_trailnow_assist_sessionsys_ai_model_configBest Practices
- ✓Start with broad content filtering policies before implementing detailed bias detection to build confidence in the system
- ✓Review AI Guardian alerts weekly during the first month to calibrate your policies and reduce false positives
- ✓Set up different monitoring levels for different AI use cases rather than applying the same strict rules everywhere
- ✓Train your help desk team on how to handle escalations when AI Guardian blocks or flags AI responses
- ✓Use AI Guardian reports to identify patterns in AI behavior rather than trying to review every individual interaction
Common Pitfalls
Creating overly restrictive policies that block too many legitimate AI responses
Start with permissive settings and gradually tighten controls based on actual problematic interactions you observe in the logs.
Expecting AI Guardian to catch every possible issue without human oversight
Use AI Guardian as an early warning system and safety net, but maintain human review processes for high-risk AI interactions.
Focusing only on content filtering while ignoring bias detection and audit capabilities
Set up basic monitoring across all AI Guardian features to understand what types of issues are most common in your environment.
Not explaining to end users why their AI responses sometimes get delayed or blocked
Create clear messaging that explains AI safety checks are in place and provide alternative ways to get help when AI responses are blocked.
Frequently Asked Questions
Is AI Guardian included in my ServiceNow license?
AI Guardian requires a separate license add-on beyond standard ServiceNow subscriptions. It's typically bundled with enterprise AI packages but check with your ServiceNow representative for specific licensing details. The cost usually depends on the number of AI interactions you expect to monitor monthly.
What AI systems does AI Guardian actually monitor?
AI Guardian works with Now Assist features, Virtual Agent chatbots, and custom AI models deployed on the ServiceNow platform. It cannot monitor external AI systems or AI tools that employees use outside of ServiceNow. The monitoring happens at the ServiceNow platform level, not at the individual AI model level.
How accurate is AI Guardian at detecting problems in AI responses?
AI Guardian's accuracy depends heavily on how well you configure your policies and rules. Content filtering for specific data types like social security numbers works very reliably, while bias detection produces more false positives and requires human judgment to interpret. Most organizations see 85-95% accuracy for content filtering and 60-80% for bias detection after proper tuning.
Does AI Guardian slow down AI response times?
AI Guardian adds typically 100-300 milliseconds to AI response times for basic content filtering. More complex bias analysis can add up to 1-2 seconds. Most users don't notice this delay, but it's something to consider for real-time chat applications where response speed is critical.
What data does AI Guardian collect about AI interactions?
AI Guardian logs the full text of AI inputs and outputs, user identifiers, timestamps, and any policy violations or flags. This data is stored within your ServiceNow instance and follows the same data governance rules as other ServiceNow records. You can configure retention periods and access controls for this audit data.
Can I customize what AI Guardian considers problematic content?
Yes, you can define custom policies for content filtering, set up industry-specific compliance rules, and adjust bias detection sensitivity. The system comes with basic templates for common use cases, but most organizations need to customize these based on their specific requirements and regulatory environment.
Do I need AI Guardian if I only use basic Now Assist features?
It depends on your organization's risk tolerance and compliance requirements. Even basic Now Assist features can potentially share sensitive information or produce biased responses. Organizations in regulated industries or those with strict data governance requirements typically implement AI Guardian regardless of which AI features they use.
Related Modules
Related Articles
Test Your Knowledge
Quick 3-question quiz — see how your ServiceNow skills stack up.
A list view on a table with millions of records is slow. Best fix?
Select an answer to continue