Monitoring conversations

The All conversations page provides visibility into all conversations that take place within Spark. It enables supervisors, administrators and support teams to analyze user interactions, understand conversation outcomes, and identify areas for improvement.

The dashboard lists all conversations that occur within Spark, and lets you break them down by outcome, type, topic and other dimensions. It allows you to monitor ongoing interactions and review completed sessions.

Accessing the All conversations page

To access the Spark overview page, navigate to Spark > All conversations from the main menu.

Filtering conversations

Use the timeframe picker in the upper-right corner to define the all conversations period. All metrics, including total conversations, outcomes, and active conversations, reflect the selected timeframe.

You can refine results using filters:

• State: In progress or Completed
• Outcome: Resolved, Escalated, Abandoned, Non-support
• Type: Incident, Request, Question, Security concern, Non-support
• Topic: The topic Spark assigned to the conversation.
• Reason: The specific reason Spark assigned for the conversation's outcome.

Refer to Monitoring Spark to understand what these filters represent and the values each one can take.

Filters apply to both the dashboard metrics and the conversation list.

Full conversations are available for up to 30 days, depending on the operational data retention settings configured for your Nexthink instance. Summarized conversation data is retained for 13 months and is available in the Spark overview dashboard and through NQL. This includes conversation state, outcome, and type.

Reviewing conversation performance

Use the dashboard sections to assess overall outcomes, identify the topics and other dimensions driving the most conversations, and spot trends before drilling into individual conversations.

Overview performance

The Overview performance section shows the number of Conversations and Active conversations for the selected period, along with the share of conversations by outcome: Resolved, Escalated, Non-support and Abandoned. A timeline chart shows how these outcomes trend over the period.

Use this section to assess overall Spark performance before drilling into a specific breakdown or conversation.

Breaking down conversations

Use the Conversations by selector to analyze conversation volume and outcomes across different dimensions. Select a breakdown from the dropdown, for example:

• Topic and Topic category: The topic and its category Spark automatically assigns to the conversation.
• State, Outcome, Type, Reason: Conversation attributes that allow for a more granular breakdown of conversation volume and outcomes.
• Channel: The communication channel where the conversation took place, such as Microsoft Teams, Nexthink Infinity or A2A.
• Operated By: Whether the employee interacted with Spark directly or another person or agent acted on the employee's behalf.
• Organization dimensions, including custom device and user classification, such as Country, Region, Entity or Department

Refer to Monitoring Spark to understand what these breakdowns represent and the values each one can take.

Use the search box within the dropdown to quickly find a breakdown dimension.

For the selected breakdown, the table shows, for each value, the number of Resolved and Escalated conversations, and the Average interactions per conversation.

Reviewing the conversation list

View all recent conversations and access details for each by selecting an item in the list.

Use the search bar above the list to filter conversations by employee or intent keywords.

The list is organized by:

• State: Displays the current progress of the conversation, such as In progress or Completed.
• Outcome: Indicates the result of the conversation, such as Resolved, Escalated, Non-support or Abandoned.
• Intent: Shows the user intent associated with the conversation, when available.
• Type: Shows the conversation classification that may be Incident, Request, Question, or Security concern. Any conversations that are not asking for IT support, as well as testing, are classified as Non-support.
• Last update: Displays the date and time of the most recent update to the conversation.
• Duration: Shows the total duration of the conversation, which is computed as the difference between the first and last message within the conversation.
• Interactions: Displays the total number of interactions exchanged between the employee and Spark.
• Employee UPN: Shows the employee unique identifier (User Principal Name).
• Topic: Shows the topic Spark assigned to the conversation, when available.

Accessing conversation details

Select a conversation from the list to open a detailed view that includes messages, outcomes, and actions.

This detailed insight helps you understand how Spark managed each conversation and provides the foundation for troubleshooting and continuous improvement.

Reviewing conversation details

Select a conversation from the conversation list to review it in more detail.

The conversation details view allows you to review:

• The messages exchanged during the conversation
• How the conversation progressed over time

AI reasoning

When relevant, the conversation details include how Spark handled the conversation and which sources it analyzed to provide an answer or execute an action. This information helps you:

• Understand how Spark leverages existing knowledge and actions to resolve employee issues.
• Review the steps taken throughout the conversation.

Reviewing conversation details, including AI reasoning when present, supports transparency and helps teams assess Spark conversations and take actions to improve knowledge or actions available to Spark.

Click on the Reasoning tab to expand the widgets and learn how Spark handled the conversation.