--- sourceDocument: Australia Enable AI sourceDocumentLink: https://www.servicenow.com/docs/r/australia/intelligent-experiences Release : - australia ft:locale : - en-US ft:publication_title : - Australia Enable AI ft:clusterId : - platai bundleId : - platai workflow : - Platform --- # Evaluated sessions # Evaluated sessions {#ariaid-title1} Release version: Australia Updated April 8, 2026 ![](https://www.servicenow.com/docs/portal-asset/ico-clock) 2 minutes to read Summarize ![AI sparkle icon](https://servicenow.com/docs/portal-asset/ai-sparkle-icon) Summarized using AI This content was generated using new OpenAI-powered functionality. Results are provided on an as is basis and are not guaranteed to be accurate or complete. ## Summary of Evaluated sessions Evaluated sessions provide ServiceNow customers with a comprehensive view of AI system performance across their portfolio, focusing on quality and safety scores. This feature enables monitoring and comparison of AI systems to quickly identify sessions that require investigation, helping maintain high standards and operational efficiency. Show full answer Show less ## Key Features * **Comparison of AI Systems:** View quality and safety scores side-by-side for each AI system through a bar chart, with options to group data by months, weeks, or days to analyze trends and assess the impact of updates or changes. * **Filtering and Sorting:** Quickly locate low-scoring sessions by sorting and filtering based on quality and safety scores, with color-coded indicators (red, orange, green) for immediate visual assessment. * **Session Details:** Access detailed session information including traces and spans to investigate issues thoroughly, understanding multi-turn interactions and task complexity. * **Performance Metrics:** Track key session metrics such as start and end times, total runtime tokens (reflecting interaction cost), and latency in milliseconds, to identify inefficiencies or performance bottlenecks. * **Role-Based Access:** Access requires the AI steward or AI asset owner roles, ensuring secure and appropriate visibility into sensitive AI system performance data. ## Using Evaluated Sessions Navigate to **All \> AI Control Tower \> Home \> Insights \> Monitor** and select the Evaluated sessions tab to begin monitoring. Use the charts and tables to: * Compare AI system performance and detect declines or improvements. * Identify sessions with low quality or safety scores for further investigation. * Correlate session performance with operational events such as deployments or outages using start and end times. * Analyze token usage and latency to optimize AI system efficiency and cost-effectiveness. ## Practical Benefits * Maintain high-quality and safe AI interactions by early detection of problematic sessions. * Optimize AI system performance by understanding behavior patterns and operational impacts. * Enhance troubleshooting capabilities with direct access to session traces and spans. View all evaluated AI sessions across your portfolio, compare quality and safety scores by AI system, and identify the sessions that need investigation. ## Key benefits {#mon-ai-evaluated-sessions-overview__key-benefits} * Compare quality and safety scores across AI systems to identify which systems are declining or performing well. * Find low-scoring sessions quickly by sorting and filtering the session list. * View score trends by AI system over weeks or months to detect patterns. * Open any session directly to investigate its traces and spans. {#mon-ai-evaluated-sessions-overview__ul_benefits} Figure 1. Evaluated sessions ## Required roles {#mon-ai-evaluated-sessions-overview__required-roles} The AI steward \[sn_ai_governance.ai_steward\] or AI asset owner \[sn_ai_asset_mgmt.ai_asset_owner\] role is required to view evaluated sessions. ## Accessing evaluated sessions {#mon-ai-evaluated-sessions-overview__accessing} View evaluated sessions by navigating to AllAI Control TowerHomeInsightsMonitor and selecting the Evaluated sessions tab. ## AI systems with evaluated sessions {#mon-ai-evaluated-sessions-overview__ai-systems-chart} Compare quality and safety scores across your AI systems using the bar chart. Each AI system displays a pair of bars representing its quality and safety scores. * Group scores by Months, Weeks, or Days to adjust the time granularity. For example, after deploying an update to an AI system, group the chart by Days to see whether the update improved scores without affecting other systems. * Select Largest score decline or Top performers from the list to change how AI systems are sorted in the chart. {#mon-ai-evaluated-sessions-overview__ul_chart_controls} ## Evaluated sessions list {#mon-ai-evaluated-sessions-overview__sessions-table} Find the sessions that need attention by reviewing the evaluated sessions table. Scores are color coded for quick scanning: red (0 to 50%), orange (51 to 75%), green (76 to 100%). Name : Session identifier. Select the name to open the session detail page and begin investigating traces and spans. See [Investigate a low-scoring session](https://www.servicenow.com/docs/fSrnmuk_wI7~xpOZ0h7L5Q "Determine the root cause of a low quality or safety score by working through the session detail, trace timeline, and span-level input and output data."). Traces : Number of traces in the session. A session with many traces represents a longer, multi-turn interaction. Low scores on sessions with few traces can indicate a fundamental issue, while low scores on sessions with many traces may point to degradation over the course of the conversation. Spans : Number of spans across all traces in the session. A high span count relative to the number of traces can indicate the AI system is taking many steps, such as tool calls, to complete its task, which can also help explain elevated latency or token usage. AI system : The AI system associated with the session. Select the name to view aggregated quality and safety scores across all evaluated sessions for that system. Quality score : Composite quality score for the session, calculated from the weighted metrics in your quality template. Sort by this column to surface the lowest-scoring sessions first. Safety score : Composite safety score for the session, calculated from the weighted metrics in your safety template. A red safety score warrants immediate investigation. Start time : When the session began. Use this to correlate low scores with known events such as deployments, configuration changes, or outages. End time : When the session ended or timed out. Total runtime tokens : Combined input and output tokens consumed by the AI system during this session. This reflects the cost of running the interaction, not the cost of evaluating it. Unusually high token counts can indicate inefficient agent behavior such as excessive retries or overly verbose responses. Latency (ms) : Combined language model response time, in milliseconds, across all traces in the session. This measures time spent in language model calls, not total session duration. Unusually high latency can point to a slow-responding model or a high number of tool-calling steps within the session. {#mon-ai-evaluated-sessions-overview__dl_sessions_columns} *[\>]: and then