Conversation Mining

The Conversation Mining feature allows you to drill down to interactions that are of interest to you or interactions that have the most potential to improve enabling you to eliminate the guesswork from manual evaluations and focus your manual efforts solely on critical interactions.

You can access Conversation Mining by navigating to Contact Center AI > Quality AI > Analyze > Conversation Mining.

The Conversation Mining has the following two sections:

• Interactions
• Audit Allocations

Note

Interactions are populated a few seconds after call termination.

Interactions

Users can see scored interactions or evaluation information at a glance from Conversation Mining. Users can apply filters to focus on specific interactions or with high potential for improvement and save the filters for auditing purposes. Interactions visible on the conversation mining screen are limited to the user's assigned queues.

The Conversation Mining interaction listing streamlines identifying specific interactions, saving time on manual reviews and enhancing decision-making and operational efficiency. You can customize the page by adding metadata and columns, improving oversight quality.

The Conversation Mining Interactions has the following key Items:

• Agents: Shows the agent who last participated in the interaction and has terminated this call. By hovering over the agents, the users can view the tagged topics and tagged intent.

• Topic Tags: Each interaction displays classified topics as tags. Hovering on this tag, shows all the relevant topics mentioned in that conversation.
  

• Intent Tags: Each interaction shows classified intents as tags. Hovering on the intent tag, shows all the relevant intents mentioned in that conversation.
  

• Actions: Allows users to assign the interaction to the desired bookmark for later reference.

  Note

  Bookmarks have to be created first from settings. For more information, see Settings for more information.

• Queues: Shows the queue in which the interaction was terminated.

  Note

  The evaluation form used to score the interaction always corresponds to the queue in which the interaction was terminated.

• Kore Evaluation Score: Shows the Kore Evaluation score (Auto QA Score) for the interaction based on the relevant evaluation form.

• Supervisor Auditor Score: Shows the Supervisor Audited score if the interaction has already been audited/manually evaluated.
• Sentiment Score: Shows the system generated sentiment score for the interaction based on the context of what was said in the interaction.
• Moments: Shows the Moments column counts for adherences, violations, and omissions related to the configured metrics of the interaction.
  

When you hover over the listed Moments, the following metrics are displayed:

• Questions Adherences: By Question Metrics that were met during the conversation.
• Violations: Speech-based violations that occurred.
• Omissions: Metrics not adhered to; including playbook steps, dialog tasks, and by question metrics.

Clicking on an interaction opens the corresponding AI assisted manual audit page that allows the user to view the conversation history and the recording.

Auditors can check the following near-miss scenarios by reviewing metrics on the audit screen:

• Evaluation Marking: If an agent's adherence is close to the required standard but not fully met, the system marks the evaluation as "No" and highlights similarities to fully adhered cases for easier comparison.

• Click-Through Navigation: The system provides clickable links (View) for near-miss agent utterances, similar to those for adhered cases, allowing for a more detailed review.

• Near-Miss Criteria: Near-miss criteria are based on predefined similarity thresholds. These thresholds help flag and navigate near-miss utterances close to adherence standards.

Columns

Allows users to filter the following default fields: Supervisor Auditor Score: This shows the Supervisor Audited score if the interaction has already been audited/manually evaluated.

• Supervisor Auditor Score: This shows the Supervisor Audited score if the interaction has already been audited/manually evaluated.

• Sentiment Score: This shows the system-generated sentiment score for the interaction based on the context of what the customer said in the interaction.

• Start Time: This shows a specified time format of the conversation in the interaction listing page (for example, 24^th May, 2024, 1:17:10 PM).

• Duration: This shows call duration (voice and chat), including talk time, hold time, and after-call work time. For example, 0h 6m 25s.

Bookmarks

Allows users to assign the interaction to a bookmark and displays all the bookmarks that a given interaction has been assigned to.

Date Range Selection

Provides the option to select the date range to the conversation interactions. Default date range selected is always the last 7 days.

Chat History

This shows all the conversation history when you click any of the agent interactions.

Filters

This provides the Filter options to filter the information based on your requirements.

Clicking any interaction will navigate you to the AI Assisted manual audit screen where you can review and evaluate the interaction.

Note

If you click any interaction that has not been assigned to you for audit, you will not be able to submit the evaluation.

Add New Filter

This new filter interaction lets you to focus on those areas of interest or with high potential for improvement, which allows users to save them for audit assignments. This helps users to filter out the options and identify which particular interaction has gone wrong.

Steps to Add New Filter:

1. Click the Filters button on the upper-right corner. The following screen appears to add a new filter.
   

The New Filter provides the following three Filter categories of interest:

Filter by Efficiency

This provides an operational view of areas of interest where there is greater potential for improvement.

1. Select a type of conversation interaction Channels, such as Chat or Voice.
   

2. Choose the Audit Status if it is Audited, Assigned, or Not Assigned.
   

3. From the Queues list, add the Queue names.
   

4. From the Agent Groups list, add the agent group name based on the queue selected.
   

   Note

   The user can filter the Agents, based on the interactions that are part of the queues and the user is part of.

5. From the Agents list, add the agent name based on the queue selected.
   

   Note

   The user can filter the Agent Groups, who are part of the queues, not based on agents in the agent group that are part of other queues.

6. Enable either of the following options:

   • Average handling time: Filters interactions based on the start and end of handling time range of interaction.
     

   • Filter by deviation from AHT: Filters interactions by % deviation from the average handling time across all interactions for the respective date range and the interactions that are going wrong.
     

   • Specify the Deviation of % number.
     

   • If No. of Transfers is selected, then specify the filters by the number of transfer that occurred within each interaction.
     

7. Click Apply to save the filter settings entered which will be stored as Unsaved Filter in the Dashboard.

   If you do not intend to use this filter to assign an audit allocation, you can apply it without saving; however, you can assign audit allocation based on filters, and can save and name it accordingly for reference during audit allocation.

8. Click Save & Apply to save the new filter with a new name.
   

9. Enable the Save Filter toggle to make the Unsaved Filter for default view in the Dashboard. All the newly created Saved Filters and Unsaved Filters will be tagged under the Saved Filters list.
   

   Note

   The filtered interactions count allows you to verify the interaction count based on the filter selections you make, this count gets dynamically recalculated as and when you update filter selections. By default, the filtered interactions count will be zero until you make the first filter selection.

Filter by Experience

Avg. Waiting Time

This provides the following filter drop down range selection conditions in seconds:

• None
• Equals
• Greater than
• Greater than equal to
• Less than
  

Sentiment Score

This indicates the positive sentiment score (higher) and negative sentiment score (lower) interactions.

Provides a slider bar to move the minimum and maximum range of interactions.

CSAT

This shows the distribution interactions across the score range that the customer has responded to the feedback service and drilled down accordingly.

Intent

This indicates the underlying cause and customer intent that the conversation pertains to.

Topic

This indicates the subject that a conversation pertains to.

Churn Monitor

This provides the underlying cause and need that a conversation relates to. It indicates the loss of customers over a specific period.

This has the following two options to churn the monitor:

Churn Risk

Provides the extent of customer churn in a given conversation. In this, the Supervisor can view the churn risk % for a given time period.

Note that the customer churn is calculated once per interaction. Customer churn is not to be calculated as a score.

Escalation

This detects the number of escalations raised to the Supervisor by a customer.

Filter by Behaviour

• Metric Name

This filter allows you to view interactions based on a specific evaluation metric, which the user can filter by selecting Pass or Fail options. The selected metric appears as a tag below the input field, which you can remove by clicking the cross button to hide the Metric Qualification field. When you access the Conversation Mining page through the Adherence Heatmap, the filter settings are automatically applied.

Metric Qualification

This filter shows the Pass or Fail options based on any queues selected in the Metric Name filter. Only evaluation metrics associated with forms in those selected queues get displayed. The filter retrieves interactions where the selected metric is relevant and part of the corresponding queue's form.

Empathy Score

This measures the level of understanding and compassion shown by the agent towards the customer situation. Provides the extent of empathy like frustration or displeasure that a customer has shown (negative sentiment). A higher score indicates a more empathetic interaction.

Crutch Word Score

This indicates the extent of filler words (for example, umm, uh, and so on) which is used by the agent. Higher score indicates the higher usage of crutch words.

Agent Playbook Adherence

This indicates the adherence percentage to the Agent AI playbook assigned to that interaction.

Kore Evaluation Score

This indicates the automated QA score associated with an interaction based on the evaluation form assigned to an interactions’s queue.

Once you Save Filter, you will get the following filters options to:

• Copy

  Allows the user to create another saved copy of the filter.

• Mark as default

  Allows the user to apply the newly created filter as a default filter whenever the call mining tab is opened.

• Edit Filter

  Allows the user to edit a saved filter.

• Delete Filter

  Allows the user to delete the saved filter.

  

Audit Allocations

This helps users to create and assign allocations to auditors for manual quality scoring.

The users can access the Audit Allocation by going to Contact Center AI > Quality AI > Analyze > Conversation Mining > Audit Allocation.

The Audit Allocations has the following options:

• Agent: Shows the name of the Auditor.
• Actions: Allows auditors to assign the allocation to the desired bookmark for later reference.
• Assigned Date: Shows the assigned date to start the audit.
• Name: Shows the audit name.
• Created By: Shows the auditor name who has initiated.
• Evaluation Form: Shows the forms list that are assigned to the QM auditors as assessments for review compliance.
• Kore Evaluation Score: Shows the Kore Evaluation score.
• Filters: Shows the filter options to search and add the filters.
• New Audit Allocation: Allows to create and assign the interactions for a new audit allocation.

New Audit Allocation

Settings

Steps to add New Audit Allocation in Settings tab:

1. Click the New Audit Allocation button. The following Settings screen appears to assign the interactions for a new audit allocation.
   

2. Under the Settings, enter a Name for the audit that needs to be done.

3. Enter a short Description of the audit which is optional.
4. Select an Evaluation Form from the drop down list to evaluate for.

5. Select Agents to search an agent from the drop down list to assign specific agents to a Queue for audit allocation.
   

6. Select Agents Groups to search an agent group from the drop down list to assign the agents group to a Queue for audit allocation.
   

7. Click Next to move to Allocation section.

Allocation

Steps to Add New Audit Allocation tab:

1. Select an Allocation Type (Random or Custom).
   1. Random allocation allows users to select randomly sampled interactions to be assigned for audit.
   2. Custom allocation allows users to select saved filters from Conversation Mining to be assigned for audit allowing focused evaluations.
2. By default, the Random radio button is selected. If you choose Random, then select a Date range.

3. Select the Channel to enable Voice toggle button, and specify the % Interactions per agent that you want to assign for audit. Based on the input, a random set of interactions is getting selected among the selected agents, and the selected queue (based on the form selection).

   a. The number of interactions per agent count below the input box displays the average number of interactions across the selected agents, which is being taken based on the % interactions per agent allocation user input.

   b. The total interactions count at the bottom of the slideout displays the total interactions, which is being selected based on random sampling and the user input across date range selection. The % interactions per agent input across channels and the count of the interactions will be assigned for this audit if needed, and this can be adjusted by altering the user input across the fields mentioned. 4. If you choose Custom, then the following screen appears to select a saved filter for Custom Allocation to assign those interactions for audits.
   

4. Select a required Filter option from the above search filter for audit.

   The total interactions count displays the total number of interactions that is being assigned for this audit based on the evaluation form (queue), agent group selection, and the filter selection.

5. Click Next to move to Assignment tab.

Assignment

Steps to Add New Audit Allocation in Assignment tab:

1. Select the Auditors from the Search filter that you want to assign interactions for manual evaluation.
   

2. Enter the % allocation of interactions that you want to allocate for each selected auditor.

   The interactions column displays the number of interactions that will be assigned for each auditor based on the allocation % input that allows you to adjust the input based on your preferences.
   

3. The total allocation percentage across all auditors must sum to 100% to enable the Create button.

Note

Once the assignment configuration is completed, such that the total allocation percentage is 100%.

1. Click Create to assign the interactions for evaluation to the selected auditors

2. The interactions that users see listed in the audit Allocation tab are the interactions that have been assigned to them for audit.

   Note that if this page is empty, it implies that no interactions are there to assign for evaluation.

3. Upon completion of evaluation for each interaction, the pertaining interaction will be removed from the audit allocation page.