Exploring Conversation Insights
Summarize
Summary of Exploring Conversation Insights
Conversation Insights in ServiceNow leverages AI to provide real-time, inferred customer satisfaction (CSAT) scores and explanatory factors based on conversation transcripts from Virtual Agent and agentic workflows. These insights help improve both automated and live agent interactions by analyzing conversations as they occur and generating actionable data without requiring direct user feedback.
Show less
Key Features
- Inferred CSAT Scores: Numerical scores from 1 (least satisfied) to 5 (most satisfied), predicted solely from conversation transcripts in real time.
- CSAT Factors: Detailed drivers of satisfaction including Resolution, Confusion, Effort, Empathy, Next Steps, Frustration, and Transfers/Escalations.
- Data Storage and Analytics: Scores and factors are stored in the Conversation Insights table [snaciinsights], with up to two years of data retention. This supports creation of custom dashboards and workflows.
- Dashboard Integration: AI Agent Analytics dashboard offers out-of-the-box visualizations of CSAT scores and factors. Insights are also accessible in Voice Agents, Now Assist Panel, and Now Assist Virtual Agent dashboards.
- Inferred CSAT by Intent: Automatically generated metrics segmented by customer intent to identify key issues and improve service for top customer concerns. Requires a sample size of at least 500 conversations for analysis.
- Comprehensive Workflow: Captures conversations from Virtual Agent and AI agents, analyzes transcripts, generates inferred CSAT scores and factors, stores insights, and makes them available for reporting and decision-making.
Key Outcomes
- Eliminates Bias and Increases Response Rates: AI-based inferred CSAT reduces reliance on traditional surveys that often suffer from low response rates and biased extreme opinions.
- Immediate Feedback: Provides real-time CSAT scores immediately after interactions, enabling faster identification of issues and trends compared to delayed survey feedback.
- Actionable Insights: CSAT factors explain the reasons behind satisfaction or dissatisfaction, guiding targeted improvements in Virtual Agent and live agent workflows.
Practical Use for ServiceNow Customers
ServiceNow customers can utilize Conversation Insights to continuously monitor and improve customer interactions across Virtual Agent and AI agent channels. The inferred CSAT scores and detailed factors help identify pain points without needing explicit customer surveys, accelerating issue detection and enhancing customer experience. Custom dashboards based on the Conversation Insights table enable tailored reporting and workflow automation to drive service improvements aligned with customer intent.
Learn how Conversation Insights can help you to augment conversation insights with AI-based Inferred customer satisfaction (CSAT) scores and factors.
Conversation Insights overview
Conversation Insights is designed to deliver Inferred CSAT scores and explanatory factors for conversations in Virtual Agent and agentic workflows. It leverages AI to analyze conversations in real time, and provides actionable insights that help improve Virtual Agent and live agent interactions, and agentic workflows.
Inferred CSAT is a numerical score from 1 (least satisfied) to 5 (most satisfied). It’s predicted entirely from conversation transcripts in real time without any input from the user. In addition to the CSAT score, the model also predicts CSAT factors that contributed to the CSAT score. The following CSAT factors are associated with the Inferred CSAT score.
- Resolution: Indicates whether the Virtual Agent or AI agent successfully resolved the user's issue without human intervention.
- Confusion: Indicates how often the Virtual Agent or AI agent misunderstood or failed to interpret the user's intent.
- Effort: Indicates the number of user turns or interactions required to reach a resolution.
- Empathy: Indicates how well the Virtual Agent or AI agent acknowledged and responded to the user's emotional tone.
- Next Steps: Captures whether the Virtual Agent or AI agent clearly communicated what the user should do next.
- Frustration: Flags signs of user dissatisfaction or repeated failed attempts during the interaction.
- Transfers and Escalations: Tracks how often the Virtual Agent or AI agent handed off the conversation to a human agent or another system.
Inferred CSAT scores and factors are calculated for each conversation. Conversational analytics applications can leverage the scores written to the Conversation Insights [sn_aci_insights] table to create custom dashboards and workflows. The AI Agent Analytics dashboard includes visualizations with Inferred CSAT scores and factors by default.
The data retention period for the Conversation Insights [sn_aci_insights] table is two years. For more information on creating custom dashboards, see Create a dashboard with the in-line editor and Select a table data source for a data visualization.
Conversation Insights are also supported by Voice Agents, Now Assist Panel and Now Assist Virtual Agent.
To view the conversation insights for Voice Agents, go to dashboard.
To view the conversation insights for Now Assist Panel or Now Assist Virtual Agent, go to .
Inferred CSAT segmented by intent
Use the additional insights introduced to report on Inferred CSAT segmented by intent. It gives better insights into why customers contact the brand and how well their needs are met. This helps companies quickly spot and fix issues with their top customer intents.
The score for Inferred CSAT segmented by intent is automatically generated when customers upgrade the Conversation Insights app. It requires a minimum sample size of 500 conversations to start and samples up to 2000 records every 15 minutes.
These metrics are available in the AI Agent Analytics dashboard, under the Insights tab, where you can find two new charts.
You can also explore the conversational insights for Now Assist Panel and Now Assist Virtual Agent. To view the dashboard, go to section or Agents section.
Additionally you can also go to dashboard to view these insights.
Conversation Insights workflow
The Conversation Insights workflow illustrates how each interaction, whether handled by Virtual Agent or an AI agent, is transformed into actionable insights. You can feed the insights directly into the dashboards for analysis and decision making. The Conversation Insights workflow shows the journey from conversations to insights on dashboards.
- Conversation sources
- Agentic AI chats
- Virtual Agent chats
- Data aggregation
- Agentic AI and Virtual Agent interactions are captured in the Conversation table [sys_cs_conversation].
- The conversation transcript, including user query, agent response, timestamps, and metadata such as the session ID and channel type are also stored in the Conversation table for processing.
- Insight generation
- The model analyzes the conversation transcript.
- Inferred CSAT scores are generated for CSAT factors such as Empathy, Resolution, Frustration, and so on.
- Insights storage
- Inferred CSAT scores and factors are stored in the Conversation Insights table [sn_aci_insights].
- The Conversation Insights table acts as a structured repository for the extracted insights.
- Dashboards
- The insights are made available to create adhoc dashboards and workflows.
- You can explore trends, performance metrics, and target improvements based on the Inferred CSAT scores.
Conversation Insights benefits
| Problem | Solution |
|---|---|
| Traditional surveys often reflect extreme opinions and low response rates. | Inferred CSAT helps solve this problem by using AI to estimate CSAT score for conversations in real time, based on full conversation transcript. This CSAT score can help to eliminate bias and reduce the need for reliance on explicit survey feedback. |
| Post-interaction feedback delays insights resulting in lagging indicators. | CSAT scores are generated immediately after the interaction, enabling faster detection of issues and trends. |
| Lack of actionable insight behind CSAT scores. | CSAT factors like Resolution, Empathy, Effort, and so on explain user satisfaction or dissatisfaction, helping you to target improvements. |