Conversation Intelligence¶

Overview¶

The Conversation Intelligence dashboard displays post-interaction analytics and provides insights following customer interactions. It gathers and presents data from those interactions, helping admins and supervisors understand agent performance, customer experience, and key interaction details.

You can filter data by date and time range and by communication type: All, Voice, or Chat. The Compare toggle highlights performance changes between the selected date range and the previous period.

The key sections of this dashboard include:

Contact Center Efficiency: Displays metrics such as Average Speed to Answer, Abandonment Rate, Customer Satisfaction Score (CSAT), and Transfer Rate.
Agent Efficacy: Shows Empathy Score, CSAT, Crutch Word Score, Agent Performance Monitor, and Agent Occupancy. The compare toggle applies here to show changes over time.
Customer Experience: Includes Average Wait Time, Net Promoter Score (NPS), Churn Risk, Sentiment Score, and Customer Churn Monitor.
Insights Mining: Provides Topic vs. Sentiment bubbles, Keyword Cloud, and Emotions.
Interaction Details: Displays session details such as Date/Time, Call ID, Agent Name/ID, Call Reason/Intent, Sentiment Score, and Dispositions.

Access Conversation Intelligence¶

Navigate to Quality AI > ANALYZE > Conversation Intelligence.

Filters¶

You can use the filters to customize the Conversation Intelligence tab.

Create a Filter¶

Steps to create a filter:

Select the Filters.
Select + Add New Filter.
On the New Filter window, choose the relevant Queues or Agents, and select Apply.

Note

Select Apply saves filters to the Unsaved Filter category for Queues or Agents, letting you review them before saving permanently.
Under the Save Filter section, enter the Filter Name.
Enable the toggle switch Make this the default view for the new filter added or save filter.
Select Save & Apply. The system displays a confirmation message after it saves the filter.

Saved Filters¶

You can view the saved filters by selecting the Filters tab.

Hovering over a saved filter displays the following options:

Duplicate
Mark as Default
Delete
Edit

Duplicate Filter¶

Steps to duplicate a filter:

Select Duplicate to copy a filter.
Enter a name in the Filter Name field for the new filter, and select SAVE.
The system displays a confirmation message and creates a copy of the filter.

Mark as Default¶

Select the Mark as Default to set the filter as the default filter.

The system displays a confirmation message when the filter sets as the default filter.

Delete Filter¶

Steps to delete a filter:

Select the Delete.
On the delete confirmation pop-up window, select Delete.

The system displays a confirmation message and deletes the filter.

Edit Filter¶

Steps to edit a filter:

Select the Edit.
Make the necessary changes, and select Save & Apply.

Clear Filters¶

This filter displays only the analytics data for the current day's agent performance, customer experience, and interaction details. When applied, it removes all previously selected date ranges and focuses solely on today's data.

Date and Time Range Selection¶

Lets you filter data by date and time range. The widget displays the current day’s data initially. Select a date and time range, and select Apply.

Compare Functionality¶

The Conversation Intelligence dashboard lets you view a comparison of metrics between a selected date range and the previous date range. When you enable the Compare toggle, the dashboard highlights changes in metrics over time. The system enables the compare toggle by default.

The dashboard displays spike and dip indicators for playbook adherence only when you enable the compare toggle. These indicators display across stages and steps for all relevant fields. An increase or decrease in a metric can have different meanings depending on context:

Green badges indicate positive changes.
Red badges indicate negative changes.

For example, if you select the last 7 days as the date range, metrics such as Average Speed to Answer shows a spike or dip representing the percentage change for June 5–12 compared to the previous period, May 29–June 5.

Channels¶

You can filter data by channels (All, Voice, Chat, and Email). This filter applies to the entire dashboard except for agent occupancy, because the system doesn't track occupancy per channel.

Contact Center Efficiency¶

Key Performance Indicators¶

Effective contact centers rely on KPIs to help managers track productivity and efficiency. In a contact center, KPIs are measurable metrics that assess how well customer service operations perform. They include:

Average Speed to Answer (Voice, Chat, Email channels): Average Speed to Answer (ASA) displays the average time an agent takes to answer inbound calls, starting when callers join the queue.

Abandonment Rate (Voice and Chat channels): The Abandonment Rate refers to the total number of customers who disconnect their calls while in the queue before reaching an agent.

Transfer Rate (Voice, Chat, and Email channels): Transfer Rate measures the percentage of customer interactions that agent or queue transfer to another resource to resolve the issue.

CSAT (Voice and Chat channels): Customer Satisfaction (CSAT) scores indicate the level of customer satisfaction with Support services that you can calculate from CSAT survey scores.

Default Zones for KPIs¶

For Average Speed to Answer:
- Voice Channel: Green up to 28 seconds, Yellow 28 to 40 seconds, Red at 40 seconds and higher.
- Chat Channel: Green up to 35 seconds, Yellow 35 seconds to 50 seconds, Red 50 seconds and higher.
For Transfer Rate (Voice and Chat Channels): Green up to 10%, and Red beyond 10%.
For Abandonment Rate (Voice and Chat Channels): Green up to 6%, and Red beyond 6%.
For CSAT: Shows the CSAT score along with the weighted value of each interaction.
- CSAT ranges from 1 to 10 (1 = lowest, 10 = highest).
- Green 8 and higher; Yellow 6 to 8; and Red 1 to 6.

The dashboard presents the KPI average for the selected date range next to the line bar and shows the percentage increase or decrease compared with the previous period. Hover over a color zone to view the call distribution by percentage and count.

Agent Efficacy¶

Empathy Score¶

Empathy plays an important role in delivering a quality customer experience. The empathy detection module analyzes both customer and agent utterances and identifies when the situation calls for empathy. It detects the sentiment in the customer’s utterance and classifies it as either empathy-seeking or non-empathy-seeking. This analyzes an agent’s utterance and classifies whether the agent is empathetic.

CSAT¶

CSAT is a widely used metric for measuring and evaluating customer satisfaction. This widget displays the average of customers' CSAT scores on a scale of 1 to 5.

Crutch Word Score¶

This score measures how frequently an agent uses crutch words during a conversation. Crutch words are filler phrases agents use to gain thinking time. Common examples include um, uh, like, you know, so, and basically. The system calculates this score from agent utterances and supports multilingual conversations by detecting subtle crutch-word patterns across languages. Learn more.

Agent Performance Monitor¶

The admins and supervisors must have a clear understanding of customer interactions and how they affect customer satisfaction. Agent Performance Monitor visualizes the relationship among key metrics such as empathy score, crutch word score, CSAT score, and sentiment score. By identifying patterns across these metrics, you can make data-driven decisions that improve the overall customer experience.

The agent performance metrics are on the Y-axis. You can select individual or multiple metrics to view and analyze. CSAT and sentiment score are the customer experience-based metrics on the X-axis.

Agent Occupancy¶

Agent occupancy represents the time agents spend handling customer interactions or performing work-related tasks such as ACW. The system shows this as a percentage.

Agent Occupancy (%) = (Total talk or chat time + Total ACW time) / (Total logged-in time) * 100

Total talk or chat time: The aggregate duration during which agents actively participate in customer interactions, including inbound and outbound calls or chats.

Total ACW time: The cumulative duration agents spend in post-call activities, including ACW, note-taking, updating customer records, or completing essential tasks linked to the interaction.

Total logged-in time: The total duration agents remain logged in to Agent AI, available to handle customer interactions.

Agent statuses include Available, Busy, Away, Break, and any custom codes the administrator configures. A pie chart visualizes the distribution of agents across these statuses, helping admins and supervisors to assess occupancy and optimize agent utilization.

Script and Playbook Adherence¶

Agents follow predefined scripts when handling customer interactions. These scripts are part of either the static playbooks configured in Agent AI or business process-specific scripts configured in the dynamic playbooks in Agent AI. Supervisors can track agents’ adherence to the configured scripts.

The adherence percentage shows as a bar graph, with each attribute represented by individual bars indicating the compliance percentage for the selected period.

The compliance percentage for each attribute appears beside the graph and shows whether compliance increased or decreased during the selected period compared with the same time range. For example, if a supervisor views compliance scores for the last week, the system shows the percentage change compared with the previous week. If the supervisor selects a one-month range, the comparison uses the previous month.

When all playbook configurations share the same steps, the system uses those steps as the default parameters for measuring adherence.

If you don't configure any Agent AI playbooks, the system uses the following default scripts (Conversation Etiquettes) to measure adherence:

Greeting: For example, Hello, My name is John Doe, and I am your customer support executive. How may I help you?
Branding: For example, Thank you for contacting Mr. John.
Privacy Policy: For example, This call gets recorded for quality and training purposes.
Hold Etiquette (Voice calls): For example, May I place you on hold for a few minutes while I pull up some information?
Customer Verification: For example, May I know your date of birth?
Proper Sign Off: For example, Thank you for reaching out. It was a pleasure to assist you. Have a great day!

Supervisors can select any playbook from the drop-down menu instead of the default options. When they choose a playbook, the system displays adherence to the specific steps and parameters defined in that playbook.

Administrators can configure the attributes for agent playbook adherence. Learn more.

Note

The Playbook Adherence tab in Conversation Intelligence is available only when Playbooks are configured in Agent AI.

Customer Experience¶

You can monitor and review historical data to assess customer experience. Analyzing these metrics helps you identify trends and make informed decisions.

The following parameters are available:

All Languages: You can select multiple languages from the All Languages drop-down, based on the languages used in the evaluation form. The system selects All languages by default. The filter shows only metrics relevant to the selected languages. When applied, language-specific interaction data appears in the Sentiment Monitor and Customer Churn Monitor widgets.
Average Wait Time: This represents the total customer wait time during a defined period (for example, one hour) divided by the number of customers served in that period.
NPS Score: The Net Promoter Score (NPS) measures customer loyalty and satisfaction. Customers rate, on a scale of 0 to 10, how likely customer recommend your product or service.
Churn Risk: Churn risk shows how many customers stop using your services during a selected timeframe. It reflects customers who move to competitors or discontinue their engagement.
Sentiment Score: The system analyzes the sentiment of each customer utterance in real-time at the keyword, phrase, or sub-topic level. After the call ends, it averages the detected sentiments to compute an overall score for the interaction, normalized on a 1 to 10 scale. This score reflects the general emotion or attitude expressed during the conversation.

Sentiment Monitor¶

The Sentiment Monitor module provides a score range from 1 to 10 for each interaction and indicates polarity as Positive, Neutral, or Negative. It determines the likely emotion involved based on the average sentiment score, such as Happy, Satisfied, or Disappointed.

The sentiment monitor graph shows the distribution of customer sentiment across different intents, enabling an admin to understand customer satisfaction and sentiment levels related to specific intents or topics of conversation.

The sentiment monitor graph is a bar chart where each bar represents an intent or topic. Each bar represents the associated sentiment across all interactions with that intent, categorizing the sentiment weightage as follows:

Positive Sentiment: Green color
Neutral Sentiment: Yellow color
Negative Sentiment: Red color

Customer Churn Monitor¶

The Customer Churn Monitor shows customer churn risk in a pie chart. It compares the churn percentage with total calls and any escalations in the same period. Hovering over a section shows its value, and selecting a section opens the calls linked to that escalation or churn category. The dashboard displays the following details:

No Customer Churn/Escalation: Number of interactions with no customer churn or escalation.
Customer Churn: Count of interactions where there has been a customer churn.
Escalation: Number of interactions where the customer has requested assistance from a supervisor or a senior.

Insights Mining¶

Insights Mining displays the top 30 topics by volume and lets you identify the topics associated with volume, sentiment, keywords and emotions for the selected date range and channel.

By default, this widget displays a blank screen without any data. Based on the date range and topic selected from the calendar, it pulls the required sentiment scores for interactions in bubbles.

All Languages¶

This lets you select multiple languages from the All Languages drop-down menu that match the languages in the evaluation form. By default, the system selects all languages. The filter shows only the metrics configured for the languages you choose.

You can filter data by channel (All, Voice, Chat, Email). This filter affects the entire dashboard, except for agent occupancy, since occupancy per channel isn't tracked.

When you select any language or channel filter, the system updates the interaction sentiment scores and bubble-plot visuals in Topics and Keyword Cloud & Emotions across all widgets.

Topics¶

This shows graded sentiment scores for interactions on a scale of 1 to 10. Each topic appears as a bubble that reflects the volume of positive, neutral, or negative sentiment. The widget also shows the top five emotions, including both positive and negative ones.

When you hover over a topic, the tooltip shows the average emotion label (for example, angry or frustrated) instead of the topic name.

Bubble Color & Sentiment Logic¶

The bubble plotting visual representation helps you identify sentiment distribution and volume for each topic. Hovering over a bubble displays a tooltip with sentiment distribution, total interactions, and average sentiment score.

Sentiment Color

Green (Positive): Indicates positive sentiment distribution and the Sentiment score is between 6 and 10.
Grey (Neutral): Indicates neutral sentiment distribution and the Sentiment score is between 4 and 6.
Red (Negative): Indicates negative sentiment distribution and the Sentiment score is between 1 and 4.
No. of Interactions: Total count of interactions for that topic.
Sentiment Score: Overall sentiment score for the topic.

The bubble size reflects the interaction volume for a topic. The system calculates this size dynamically within the selected date range and scales it to show the topic’s volume relative to the minimum and maximum values. By default, all the three selected sentiment colors, give you a combined view of all topics across sentiments. You can select a specific sentiment color to view the most relevant topics in that category.

The color logic for sentiment:

1 ≤ x < 4: Negative (Red)
4 ≤ x < 6: Neutral (Grey)
6 ≤ x < 10: Positive (Green)

Note

Only the top 30 most significant topics (based on interaction volume) are displayed on the widget.

For example, the Payment bubble indicates sentiment distribution as Positive 68% (green), Neutral 10% (gray), and Negative 22% (red).

The bubble’s circumference represents these percentages. When you hover over the Payment bubble, the tooltip shows the average emotion index, such as Flexible, Accurate, or Fast, based on the average sentiment score.

Keyword Cloud & Emotions for All¶

This widget shows relevant keywords for the selected Topic from the previous widget. It excludes stop words and other common terms.

Keyword Search¶

Supervisors use Keyword Search to find and analyze keywords across interactions by topic.

Select one topic or All Topics and enter a keyword to search interactions.
Hover over a keyword to view total mentions and unique interactions.
Select a topic bubble or choose a topic from the drop-down list to filter related keywords.
Select All Topics to view keywords and interaction details across all topics.
When a keyword appears in multiple topics, the system reflects topic-specific relevance.
The keyword cloud shows only the searched keyword and available semantic variations.
Select a semantic variation to view its related interactions in the Interaction Details panel.
When you select a semantic variant in the keyword cloud, the system displays the interactions related to that specific keyword in the Interaction Details panel.

Interaction Details¶

The Interaction Details widget shows information based on default settings, which you can configure. When you select specific topics or keywords, the widget displays the relevant information. If you make no selection, it shows data based on the highest sentiment score or other settings you configure.

This section includes the following fields:

Date/Time of the session
Call ID
Agent name/ID (you can select from a dropdown list of agent groups/agents)
Description (text reference to the keyword cloud)
Call Reason/Intent
Sentiment Score (configurable from high to low)
Dispositions

Send Feedback