Voice page in assistant analytics
Summarize
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 Voice page in assistant analytics
The Voice page in Assistant analytics within Assistant Designer enables ServiceNow customers to monitor and analyze the performance of their voice assistants. It organizes key performance metrics across four tabs—Overview, Performance, Insights, and Assist consumption—to provide a comprehensive view of voice assistant interactions, resolution effectiveness, tool usage, authentication, conversation quality, and AI agent assist consumption.
Show less
Customers can filter data by voice assistant, language, communication channel, and date range to tailor insights to specific contexts. Metrics tagged as AI Inferred are derived from large language model (LLM) transcript analysis and should be reviewed carefully before making decisions.
Overview tab
- Total voice conversations: Tracks the total number of voice interactions over a selected period to identify usage trends and set performance benchmarks.
- Resolution rate (AI Inferred): Shows the percentage of conversations resolved by the AI without human intervention, useful for evaluating assistant effectiveness.
- Live agent transfer rate: Measures how often conversations are escalated to live agents, helping identify and reduce unnecessary transfers.
- Average handle duration: Indicates the average time the voice assistant spends handling conversations, highlighting opportunities to streamline interactions.
- Average turns: Reflects conversational efficiency by showing the average number of exchanges per conversation.
- Average CSAT score: Provides customer satisfaction scores gathered from post-conversation surveys to track user sentiment over time.
- AI voice agent performance: Displays conversation counts and AI-inferred resolution rates per voice agent to identify high and low performers.
Performance tab
- Tool execution metrics: Presents execution counts and performance times (median, 90th, and 99th percentiles) for various tools invoked during voice conversations, such as flow actions and RAG-based searches, enabling optimization of frequently used tools.
- Guest authentication performance: Shows authentication times and outcomes to assess and improve guest verification efficiency during voice interactions.
- Response time metrics: Provides response time percentiles indicating how quickly the voice assistant responds after user input, essential for ensuring timely user experiences.
Insights tab
- Additional conversation outcomes: Tracks events like immediate live transfers, early disconnects, session expirations, and ticket creation to understand conversation flow and issues beyond resolution rates.
- User effort required: Measures user effort levels during conversations based on signals like transfers and wait times, highlighting areas to reduce friction.
- AI agent empathy: Assesses the politeness and attentiveness of AI agents, informing improvements in conversational tone and user engagement.
- User/AI confusion and frustration: Detects misunderstandings and user dissatisfaction during conversations to identify and address pain points.
Assist consumption tab
This tab provides insights into AI agent assist usage, where an assist records each voice agent action during conversations. Assists are categorized into Small, Medium, and Large tiers based on the number of actions performed.
- Total assists consumed: Tracks the volume of AI agent actions over time to monitor assist usage trends.
- Assists by tier: Breaks down assists by size to understand the complexity of agent actions.
- Assist consumption by agent: Shows average actions per call and total assists per AI voice agent, helping identify which agents perform more complex or frequent actions.
Practical Benefits for ServiceNow Customers
Using the Voice page in Assistant analytics, customers can:
- Gain detailed visibility into voice assistant performance and user interactions.
- Identify trends and patterns in conversation volume, resolution rates, and escalations.
- Optimize tool usage and reduce response and authentication times.
- Monitor conversational quality signals such as empathy, user effort, and frustration to enhance user experience.
- Track AI agent assist consumption to manage and optimize resource usage effectively.
Overall, this enables continuous improvement of voice assistants to deliver efficient, satisfying, and scalable voice interactions aligned with business goals.
Monitor the performance of voice assistants from the Voice page of Assistant analytics in Assistant Designer.
The Voice page shows performance metrics for your voice assistants, organised across four tabs: Overview, Performance, Insights, and Assist consumption. Use these tabs to monitor conversation volume, resolution rates, tool execution, authentication performance, conversation quality signals, and AI agent assist usage.
You can filter the data on all tabs using the following options:
- Filter by voice assistant: View metrics for specific voice assistants.
- Filter by language: View metrics by conversation language, such as English, German, Spanish, and so on.
- Filter by communication channel type: View metrics by channel, such as phone, mobile app - iOS, mobile app - Android, or web browser.
- Filter by date range: View metrics for a specific time period.
Scorecard widgets display a trend indicator below the metric value, showing the change in value compared to the equivalent previous period. The indicator includes the absolute change, the percentage change, and the comparison date range.
Note:
Metrics marked with the AI Inferred tag are calculated through large
language model (LLM) transcript analysis. Results may vary and may not be fully accurate.
Review AI Inferred metrics before acting on them.
Overview tab
The Overview tab shows high-level conversation volume, resolution performance, and AI voice agent activity for the selected date range.
- Total voice conversations
- This area of the dashboard shows the total number of conversations with voice assistants in
the selected date range. Use this metric to track growth in voice interactions and set
benchmarks for assistant performance. Select Related records to view
the underlying records.
Figure 2. Total voice conversations - Total voice conversations over time
- This area of the dashboard shows how conversation volume has changed over the selected date
range. Use this chart to identify patterns such as peaks and dips in usage and correlate them
with changes to your assistant configuration. Hover over a date to view the total number of
conversations for that date.
Figure 3. Total voice conversations over time - Resolution rate (AI Inferred)
- This area of the dashboard shows the percentage of conversations resolved by the voice AI,
with no human follow-up needed. This is measured through large language model (LLM) transcript
analysis. Select Related records to view the underlying records.
Figure 4. Resolution rate - Resolution rate over time (AI Inferred)
- This area of the dashboard shows how the resolution rate has changed over the selected date
range. Use this chart to track the impact of assistant updates and configuration changes on
resolution performance. Hover over a date to view the resolution rate percentage for that
date.
Figure 5. Resolution rate over time - Live agent transfer rate
- This area of the dashboard shows the percentage of conversations transferred from the voice
assistant to a live agent in the selected date range. Use this metric to identify common
escalation triggers and optimise assistant workflows to reduce unnecessary transfers. Select
Related records to view the underlying records.
Figure 6. Live agent transfer rate - Live agent transfer rate over time
- This area of the dashboard shows how the live agent transfer rate has changed over the
selected date range. Use this chart to identify periods of increased escalation and correlate
them with conversation patterns or assistant configuration changes. Hover over a date to view
the transfer rate percentage for that date.
Figure 7. Live agent transfer rate over time - Average handle duration (with voice assistant)
- This area of the dashboard shows the average duration of the AI-handled portion of voice
conversations, excluding time spent with a live agent after transfer. Use this metric to
identify opportunities to streamline conversations and reduce resolution time.
Figure 8. Average handle duration (with voice assistant) - Average handle duration (with voice assistant) over time
- This area of the dashboard shows how the average handle duration has changed over the
selected date range. Use this chart to track whether workflow optimisations are reducing the
time the voice assistant spends handling conversations. Hover over a date to view the average
handle duration in seconds for that date.
Figure 9. Average handle duration (with voice assistant) over time - Average turns
- This area of the dashboard shows the average number of turns taken by the voice assistant
per voice conversation in the selected date range. Use this metric to
assess how efficiently the assistant resolves user requests.
Figure 10. Average turns - Average turns over time
- This area of the dashboard shows how the average number of conversational turns has changed
over the selected date range. Use this chart to identify whether conversations are becoming
more or less efficient over time. Hover over a date to view the average turn count for that
date.
Figure 11. Average turns over time - Average CSAT score
- This area of the dashboard shows the average customer satisfaction (CSAT) score for voice
conversations in the selected date range. CSAT scores are captured through your own
post-conversation survey solution. Select Related records to view the
underlying records.
Figure 12. Average CSAT score - Average CSAT score over time
- This area of the dashboard shows how the average CSAT score has changed over the selected
date range. Use this chart to evaluate the impact of assistant updates on user satisfaction
over time. Hover over a date to view the average CSAT score for that date.
Figure 13. Average CSAT score over time - AI voice agent performance
- This area of the dashboard shows the conversation count and resolution rate for each AI
voice agent invoked during conversations. The resolution rate column is AI Inferred. Use this
table to identify high-performing agents and improve underperforming ones.
The table includes the following columns:
- Voice AI agent — name of the AI voice agent.
- Conversations — total number of conversations handled by the agent.
- Conversation change — change in conversation count compared to the previous period.
- Resolution rate (%) — percentage of conversations resolved by the agent, measured through LLM transcript analysis (AI Inferred).
Figure 14. AI voice agent performance
Performance tab
The Performance tab shows tool execution metrics, authentication performance, and response time data.
- Execution count by tool type
- This area of the dashboard shows the number of tool executions categorised by type,
including flow actions and RAG-based search, for the selected date range. Use this metric to
understand which tools are most frequently invoked during voice conversations. See Add tools and information to an AI agent for information on tool types.
Figure 15. Execution count by tool type - Tool execution time (50th percentile)
- This area of the dashboard shows the tool execution time within which 50% of tool
executions were completed. Tool time is the time required for the voice agent to complete
actions using tools. Select Related records to view the underlying
records.
Figure 16. Tool execution time (50th percentile) - Tool execution time (90th percentile)
- This area of the dashboard shows the tool execution time within which 90% of tool
executions were completed. Only 10% of executions took longer than this time. Select
Related records to view the underlying records.
Figure 17. Tool execution time (90th percentile) - Tool execution time (99th percentile)
- This area of the dashboard shows the tool execution time within which 99% of tool
executions were completed. Only 1% of executions took longer than this time. Select
Related records to view the underlying records.
Figure 18. Tool execution time (99th percentile) - Tool Performance
- This area of the dashboard shows performance broken down by individual tool. Success rate
reflects the share of executions that completed without error. P50, P90, and P99 columns
represent execution time percentiles in seconds.
The table includes the following columns:
- Type — the tool type, such as Flow actions or RAG-based search.
- Tool — the name of the specific tool invoked.
- Success rate — the percentage of executions that completed without error.
- P50 Time (s) — median execution time in seconds.
- P90 Time (s) — 90th percentile execution time in seconds.
- P99 Time (s) — 99th percentile execution time in seconds.
Figure 19. Tool Performance - Guest authentication time (50th percentile)
- This area of the dashboard shows the time within which 50% of guest authentication
sequences were completed, measured from the first voice AI identification question until the
authentication API call response. Select Related records to view the
underlying records.
Figure 20. Guest authentication time (50th percentile) - Guest authentication time (90th percentile)
- This area of the dashboard shows the time within which 90% of guest authentication
sequences were completed. Only 10% took longer than this time. Select Related
records to view the underlying records.
Figure 21. Guest authentication time (90th percentile) - Guest authentication time (99th percentile)
- This area of the dashboard shows the time within which 99% of guest authentication
sequences were completed. Only 1% took longer than this time. Select Related
records to view the underlying records.
Figure 22. Guest authentication time (99th percentile) - Guest Authentication Outcomes
- This area of the dashboard shows the number of interactions by authentication outcome.
Successful means the caller was authenticated at any point during the interaction. Failed means
every authentication attempt was unsuccessful.
Figure 23. Guest Authentication Outcomes - 50th percentile response time
- This area of the dashboard shows the response time within which 50% of voice assistant
responses were completed, measured from when the user finishes speaking to when the voice
assistant responds. Only 50% of responses took longer than this time.
Figure 24. 50th percentile response time - 90th percentile response time
- This area of the dashboard shows the response time within which 90% of voice assistant
responses were completed. Only 10% of responses took longer than this time.
Figure 25. 90th percentile response time - 99th percentile response time
- This area of the dashboard shows the response time within which 99% of voice assistant
responses were completed. Only 1% of responses took longer than this time.
Figure 26. 99th percentile response time
Insights tab
The Insights tab shows conversation outcome data and key sentiment signals measured through large language model (LLM) transcript analysis.
- Additional conversation outcomes
- This area of the dashboard shows conversation outcomes beyond resolution and live agent
transfer. Immediate outcomes refer to events occurring within the first 30 seconds of the
call.
The table includes the following columns: Outcome, Conversation count, Conversation change, and Percent of total conversations.
Outcomes tracked include:
- Immediate live transfers — conversations where the caller requested transfer to a live agent within the first 30 seconds.
- Immediate disconnects — conversations that disconnected within the first 30 seconds.
- Session expiration — conversations that ended due to session timeout.
- Ticket created — conversations that resulted in a ticket, such as an incident or case record, created for follow-up.
Figure 27. Additional conversation outcomes - User effort required with AI agents
- This area of the dashboard tracks how much effort a user had to put in during a
conversation, based on signals like transfers, wait times, and escalations. Scored as Low,
Medium, or High. This is measured through large language model (LLM) transcript analysis.
Select Related records to view the underlying records.
Figure 28. User effort required with AI agents - AI agent empathy
- This area of the dashboard measures how politely and attentively the agent acknowledged the
user's needs and concerns throughout the conversation. Scored as Low, Medium, or High. This is
measured through large language model (LLM) transcript analysis. Select
Related records to view the underlying records.
Figure 29. AI agent empathy - User / AI agent confusion
- This area of the dashboard indicates whether the user and agent misunderstood each other at
any point in the conversation. Scored as Yes or No. This is measured through large language
model (LLM) transcript analysis. Select Related records to view the
underlying records.
Figure 30. User / AI agent confusion - User frustration with AI agents
- This area of the dashboard indicates whether the user expressed frustration, through
complaints, sarcasm, or dissatisfaction, during the conversation. Scored as Yes or No. This is
measured through large language model (LLM) transcript analysis. Select
Related records to view the underlying records.
Figure 31. User frustration with AI agents
Assist consumption tab
Note:
The data on this tab is for informational purposes only and should not be relied upon as a
definitive statement of your Now Assist usage for billing purposes.
The Assist consumption tab shows AI agent assist usage metrics for the selected date range. An assist is recorded each time an AI voice agent completes an action during a conversation. Assists are categorised into three tiers based on the number of actions taken by the voice agent: Small, Medium, and Large.
- Total assists consumed
- This area of the dashboard shows the total number of assists consumed by AI voice agents in
the selected date range. Select Related records to view the underlying
records.
Figure 32. Total assists consumed - Total assists consumed over time
- This area of the dashboard shows how the total number of assists consumed by AI voice
agents has changed over the selected date range. Hover over a date to view the number of
assists consumed for that date.
Figure 33. Total assists consumed over time - Total assists consumed by tier
- This area of the dashboard shows the total assists consumed, broken down by tier: Small,
Medium, and Large.
Figure 34. Total assists consumed by tier - Assist consumption by agent
- This area of the dashboard shows the average actions per call and total assists consumed,
broken down by AI voice agent. Assist tier (small, medium, large) is based on the number of
actions taken by the voice agent.
The table includes the following columns:
- AI voice agent — name of the AI voice agent.
- Average actions per call — average number of actions the agent took per conversation.
- Assists Consumed — total number of assists consumed by the agent.
Figure 35. Assist consumption by agent