Introduction

Agentic Analytics is an advanced AI-powered analytics tool in wolkvox Manager that allows you to analyze customer conversations, evaluate AI agent performance, and identify opportunities for operational improvement.

This functionality transforms information obtained from conversations into indicators, statistics, summaries, and recommendations that facilitate service quality monitoring and the optimization of customer service processes.

In addition to displaying information through interactive dashboards, Agentic Analytics includes an artificial intelligence assistant that allows you to query your operation's data using natural language. It also provides a weekly summary of key findings and an AI-generated podcast so you can listen to the analysis results.

What Is Agentic Analytics Used For?

• Evaluate AI agent performance: review indicators such as FCR, CSAT, NPS, and AHT to assess conversation effectiveness.
• Identify opportunities for improvement: detect recurring problems, difficulties during conversations, and areas that can be optimized.
• Analyze customer experience: understand conversation sentiment, customer satisfaction levels, and reasons for contact.
• Monitor operational performance: visualize conversation volume, channel distribution, and the most frequent categories.
• Review individual conversations: examine conversation history, listen to call recordings, and view analysis results.
• Obtain AI-powered insights: ask questions about operational data and receive answers, diagnoses, and recommendations.
• Access periodic summaries: review the operation's key findings and listen to a weekly podcast generated by artificial intelligence.

Activate Agentic Analytics

To start using Agentic Analytics, you must configure which channels you want to analyze and what percentage of their conversations will be included in the analysis.

1. Click the "Settings" button, identified by the gear icon located in the upper-right menu.
2. Select the "Agentic Analytics" tab.
3. You will find two configuration options:
   • Chat Analytics
     • Enable Chat Analytics: enables AI-powered analysis of chat conversations.
     • Percentage to Analyze %: allows you to define the percentage of chat conversations to be analyzed.
   • Voice Analytics
     1. Enable Voice Analytics: enables AI-powered analysis of voice conversations.
     2. Percentage to Analyze %: allows you to define the percentage of calls to be analyzed.
   • Select the checkboxes corresponding to the channels you want to analyze.
   • Enter the desired analysis percentage for each enabled channel.
4. Click the "Save" button to apply the configuration.

Important: analysis percentages are configured independently for voice and chat channels. For example, you can configure 50% for Chat Analytics and 100% for Voice Analytics, depending on the needs of your operation.

Access Agentic Analytics

Once the service has been configured, you can view the results from the dashboard section of wolkvox Manager.

1. In the left sidebar, go to "Dashboard". 
2. Select the "Agentic Analysis" tab. In this section, you will find the "Real Time", "Information", "Interaction Analytics", and "Recommendations" tabs, where you can access the different analysis features.

View General Operational Information

In the "Information" tab, you will find a summary of the main aspects identified by Agentic Analytics.

This section organizes the information into the following categories:

• Objective
  • Describes the general purpose of the virtual assistant and the activities it performs within the operation, such as answering inquiries, providing information, managing requests, or executing automated processes.
• Opportunities for Improvement
  • Displays problems and optimization opportunities detected by artificial intelligence based on conversation analysis.
  • Depending on the findings, recommendations may include information about:
    • Difficulties understanding customer requests.
    • Incorrect responses or difficulties executing processes.
    • Requests that were not satisfactorily resolved.
    • Errors in validations or automated procedures.
    • Situations that cause customer frustration or dissatisfaction.
  • Identified opportunities may include a description of the problem, conversation examples, possible causes, impact, and recommendations for improving AI agent behavior.
• Podcast
  • Agentic Analytics allows you to listen to an AI-generated podcast summarizing the main findings obtained from conversations during the previous week.
  • To use it, click the "Play Audio" button.

During playback, the following controls will be available:

• Pause: temporarily stop playback.
• Resume: continue listening from where playback was paused.
• Stop: end audio playback.
• Download: save the podcast to your device.

This functionality allows you to learn about key results and opportunities for improvement without having to manually review each conversation.

"Real Time" Tab

The "Real Time" tab in Agentic Analytics allows you to monitor active conversations handled by AI agents across voice and chat channels. From this section, you can view performance indicators, identify interactions that present risks, and monitor the progress of each conversation as it happens.

 

General Indicators

Three indicators appear at the top of the section to provide an overview of active conversations:

Indicator	Description
Active Interactions	Displays the total number of active conversations, distinguishing how many belong to the voice channel and how many to the chat channel.
At-Risk Interactions	Indicates the number of active conversations classified as high or critical risk, helping identify situations that require attention.
Average Sentiment	Displays the average sentiment value identified in active conversations, providing an overview of customer sentiment during service interactions.

 

Monitoring Active Interactions

Below the indicators, a table displays the active conversations handled by AI agents.

To facilitate monitoring, this section includes a search field for agents or customers, filters to display all conversations or only those from the Voice or Chat channels, and a sorting selector.

The table contains the following fields:

Field	Description
AI Agent	Name and identifier of the artificial intelligence agent handling the conversation.
Customer	Name and identification information of the customer participating in the conversation.
Channel	Identifies whether the interaction is taking place through voice or chat.
Latency	Displays the latency recorded during the interaction, expressed in milliseconds (ms), including the most recently recorded value.
Dur.	Time elapsed since the active conversation began, expressed in hours, minutes, and seconds, as applicable.
Sentiment	Classification of the sentiment identified during the conversation, such as neutral.
Risk	Risk level identified during the interaction, such as low.
Comment	Displays observations related to the progress of the conversation.

 

 

 

How to View Real-Time Conversation Details?

1. Locate the conversation you want to monitor in the active interactions table. You can use the search field or channel filters. Click the corresponding record.
2. A right-side panel will open with detailed information about the selected conversation.

At the top of this panel, you will find the AI agent's name and identifier, customer information, communication channel, latency, conversation duration, identified sentiment, and risk level.

 

 

Below this information, you will find the following tabs:

 

Live Transcription

Allows you to view the conversation content as it unfolds, distinguishing messages from the ASSISTANT and the CUSTOMER.

This functionality makes it easier to monitor the responses generated by the AI agent and the requests made by the customer without waiting for the conversation to end.

At the bottom of the panel, you will also find the "Live Summary" section, which provides a description of what has occurred during the interaction up to that point. This allows you to quickly understand the service context without having to review the entire transcript.

 

 

Timeline

Provides a chronological view of the interaction's progress, complementing the information available in the live transcription.

 

 

Benefits of Real-Time Monitoring

• Operational visibility: allows you to see how many conversations AI agents are handling and through which channels.
• Risk identification: helps detect interactions classified as high or critical risk.
• Customer experience monitoring: allows you to review the sentiment identified during active conversations.
• AI response monitoring: makes it easier to review what customers say and how AI agents respond in real time.
• Quick access to context: provides summaries of ongoing conversations to understand their progress and identify potential opportunities for improvement.

View Interaction Analytics

Go to the "Interaction Analytics" tab to view statistics and results from conversations analyzed through Agentic Analytics.

At the top, you will find six main indicators that allow you to evaluate operational performance.

Each indicator allows you to compare results with the previous period. Depending on the selected date range, trend charts may also be displayed to help identify performance variations.

Important: in this dashboard, the term "Interactions" refers to all communications included in the analytics, not only emails or social media comments, which wolkvox classifies as interactions within its omnichannel management.

Select the Analysis Period

In the upper-right corner of this section, you will find a dropdown menu with the following options:

• Last Day
• Last 7 Days
• Last 30 Days
• Last 90 Days

Select the period you want to review. The indicators and statistics in this section will update according to the selected range.

This allows you to compare operational performance over the short, medium, and long term.

View Details of Each Conversation

On the left side of "Interaction Analytics", you will find the "Interaction Details" panel, which displays information about a specific conversation and the results obtained through artificial intelligence analysis.

In this panel, you can view:

Below this information, you will find two tabs:

Summary

Displays an AI-generated summary of the main events that occurred during the conversation.

It also includes a justification explaining the evaluation result, considering what the customer expressed and how the interaction was handled.

Conversation

Allows you to review the conversation content, displaying the messages exchanged between the customer and the AI agent.

This makes it easier to identify what information the customer provided, how the agent responded, and the outcome of the interaction.

Play Call Recordings

When the conversation belongs to the voice channel and its recording is available, you will find an audio player that allows you to listen to the call.

To download it, click the three vertical dots button in the player and select "Download".

Navigate Between Conversations

At the top of the panel, you will find the following buttons:

• Previous: allows you to view the previous conversation.
• Next: allows you to view the next conversation.

You can also select a conversation directly from the "Last 100 Interactions" chart to view its details in this panel.

Analyze Statistical Charts

In the central area of the "Interaction Analytics" tab, you will find different charts that provide an overview of conversation behavior.

Last 100 Interactions

This chart displays the results of the last 100 interactions available in the analysis.

In the upper-right corner, you will find a dropdown menu with four criteria:

Each point on the chart represents a conversation.

To view conversation details:

• Select the analysis criterion you want to display.
• Locate the point corresponding to the conversation.
• Click the point.
• The information for the selected conversation will automatically appear in the "Interaction Details" panel on the left.

This allows you to identify conversations with negative results, low satisfaction scores, or requests that were not resolved during the first contact and review what happened during the interaction.

Conversations by Category

Displays a pie chart that distributes conversations according to the identified reason or category.

It allows you to view the percentage represented by each category within the total analyzed, such as general inquiries, information requests, technical issues, transfers, or complaints and claims.

This information helps identify the most frequent inquiry topics and the processes that account for the highest conversation volume.

Conversations by Channel

Displays a pie chart showing the percentage distribution of conversations by channel.

The visualization identifies conversations corresponding to Chat and Voice, allowing you to understand each channel's contribution to the total analyzed volume.

Traffic in the Last 24 Hours

Displays a bar chart showing conversation volume trends over the last 24 hours.

The bars distinguish voice and chat conversations, making it easier to identify periods of higher activity and compare traffic between both channels.

Query Operational Data Using the Agentic Analytics Assistant

In the right-side panel, you will find Agentic Analytics, an AI-powered conversational assistant that allows you to query the operation's analyzed information.

With this tool, you can obtain summaries and diagnoses, identify recurring problems, and request recommendations using natural language questions.

How to Submit a Query?

• Locate the Agentic Analytics panel on the right side of the dashboard.
• In the input field at the bottom, enter the question or instruction you want to submit. You will also find a microphone button for entering instructions by voice.
• Submit your query.
• Agentic Analytics will analyze the available information and generate a response related to the operation's data.
• You can ask follow-up questions within the same conversation to explore the results in greater detail.

Example Queries:

• "What have been the main operational issues today?"
• "What opportunities for improvement have been identified in our AI agents?"
• "What are the most frequent inquiry categories?"
• "What issues are affecting the operation's FCR?"
• "Identify the conversations with the lowest customer satisfaction scores and explain the possible causes."
• "Give me a summary of AI agent performance over the last 7 days."
• "What recommendations do you have for reducing average handling time?"

Example of a Generated Analysis

If you ask about the operation's main problems, Agentic Analytics can provide a diagnosis that includes indicators, channel distribution, the most frequent categories, detected issues, and opportunities for improvement.

Responses are generated based on the information available for analysis, so their content depends on the data recorded and processed.

View Conversation History

• Click the menu button, identified by three horizontal lines, located in the upper-left corner of the panel.
• A list of previous conversations with Agentic Analytics will appear.
• Select a conversation to review its content again.

1. Download a Conversation
   1. Open the conversation you want to download.
   2. Click the button with the cloud and downward arrow icon, located in the upper-right corner of the panel.
   3. The downloaded conversation will open in a tab of your operating system's default web browser.
2. Start a New Conversation
   1. Click the button with the broom icon, located in the upper-right corner of the panel.
   2. A new conversation will begin, allowing you to submit other queries without continuing the previous exchange.
   3. These tools allow you to organize queries, preserve analysis results, and return to previous conversations when necessary.

"Recommendations" Tab

The "Recommendations" tab in "Agentic Analysis" allows you to identify opportunities to improve AI agents across voice and chat channels based on the analysis of customer conversations.

Agentic Analytics analyzes interactions in which problems were identified during customer service and generates specific recommendations to improve the behavior, instructions, and logic of flows configured in wolkvox Studio.

These recommendations can be implemented using Agentic Flow, the artificial intelligence tool in wolkvox Studio that allows you to modify routing points through natural language instructions.

When Are Recommendations Generated?

Recommendations are generated for voice or chat AI agents that record more than 100 interactions with negative sentiment or an FCR value of 0 during the previous week.

When this condition is met, Agentic Analytics analyzes the corresponding interactions and identifies recurring problems that can be corrected by adjusting the service flow.

How to View Agentic Analytics Recommendations?

1. Log in to wolkvox Manager and select Dashboard from the left sidebar.
2. Go to the "Agentic Analysis" tab.
3. Select the "Recommendations" tab.

1. In the left panel, you will find the generated recommendations along with the number of pending recommendations. Each recommendation displays the related routing point, a title describing the proposed improvement, the corresponding channel (Chat or Voice), and a classification based on the type of modification, such as Prompt, Logic, or Tools. Click the "Review" button for the recommendation you want to view.
2. When you select a recommendation, its detailed information will appear in the central panel.

What Information Does Each Recommendation Contain?

Each recommendation includes the following elements:

• Problem: description of the situation identified during conversations. Explains the behavior affecting customer service, such as incorrect responses, failed validations, or difficulties completing a request.
• Impact: quantitative information about interactions related to the detected problem.
• Recommendation: proposed modification to correct the identified behavior. It may include adjustments to agent instructions, flow conditions, validations, or components.
• Supporting Interaction: list of conversations used as evidence to support the recommendation.
• Routing Point: identifier and name of the routing point associated with the recommendation.
• Mode: indicates the mode associated with the analyzed flow.

How to Review Conversations Supporting a Recommendation?

In the "Supporting Interaction" section, you will find the conversations in which situations related to the problem were identified.

To review a conversation:

1. Select one of the records from the supporting interactions list. A right-side panel called "Interaction Details" will open. In this panel, you can view information such as the customer, phone number, conversation identifier, date and time, duration, channel, category, FCR, transfer to a human agent, CSAT, and identified sentiment.
   • Select the "Summary" tab to view a description of what occurred during the interaction and its justification.
   • Select the "Conversation" tab to review the messages exchanged between the customer and the AI agent.
2. This information allows you to verify the context in which the problem occurred and determine whether the proposed modification is appropriate for the operation.

How to Apply a Recommendation in wolkvox Studio?

After reviewing the recommendation and its supporting conversations, you can use it to modify the corresponding routing point.

1. Select the recommendation you want to implement. 
   • Review the identified problem, its impact, and the proposed modification. 
   • Click the "Apply" button, located in the lower-right corner of the recommendation panel.

Important: clicking "Apply" does not automatically execute the recommendation. The system opens the routing point and prepares the instruction in Agentic Flow. The administrator must press Enter to start executing the changes.

1. The system will automatically redirect you to wolkvox Studio and open the associated routing point.
   • Agentic Flow will open with the instruction corresponding to the recommendation already loaded in its input field.
   • Review the instruction and press Enter to submit it to Agentic Flow.
   • Agentic Flow will process the request and make the corresponding modifications to the flow.
   • Review the changes made, verify compilation, and perform tests to confirm that the routing point behaves as expected.

How to Dismiss a Recommendation?

If, after reviewing the information, you determine that a recommendation is not applicable or does not meet the needs of your operation, you can use the "Delete" button located in the lower-left corner of the details panel.

This allows you to dismiss recommendations you do not want to implement and focus your review on those that add value to customer service processes.

Benefits of Agentic Analytics Recommendations

• Continuous improvement of AI agents: helps identify and correct recurring problems in conversations.
• Recommendations supported by real interactions: makes it easier to validate opportunities for improvement through conversations in which difficulties occurred.
• First Contact Resolution optimization: helps identify behaviors that affect FCR and propose adjustments to improve AI agents' resolution capabilities.
• Reduction of negative experiences: helps detect situations that cause customer frustration or dissatisfaction.
• Integration with Agentic Flow: facilitates the implementation of proposed improvements directly in wolkvox Studio routing points.
• Greater control over changes: administrators can review, apply, or dismiss each recommendation before modifying customer service flows.
• In this way, Agentic Analytics connects conversation analysis with the optimization of AI-powered flows, facilitating the identification of problems and the implementation of improvements in customer service.