Agent analytics show credible usage reports that can be reviewed by upper management and executives.
They provide a comprehensive analytics layer that enables organizations to monitor, understand, and optimize the performance of AI agents across their employee support ecosystem.
AI effectiveness analytics allows you to:
- Monitor AI effectiveness
- Identify improvement opportunities
- Create clear metrics on AI conversations and activity
- Have visibility into AI resolution effectiveness
- Gain insight into employee feedback on AI responses
- Compare agent-level performance
- Analyze operational insights across channels, tools, and topics
Use case
Problem
Without performance analytics, AI adoption becomes difficult for HR professionals to trust, evaluate, and optimize.
Solution
This feature provides you with the insights needed to monitor AI effectiveness, identify improvement opportunities, and apply them successfully.
Ultimately, this strengthens Tier-0 support automation and enables confident expansion of AI-driven service delivery.
Agent analytics blocks
The blocks displayed on this dashboard are preset and defined to provide you with the best overall Agent analytics experience.
The agent analytics metrics are as follows:
Conversations
Any conversation sessions, defined as conversations with at least one message sent within the period, are captured automatically and displayed here in a graph with a trend over time.
Message
Here, the count of messages sent by AI agents within the selected period is automatically captured and displayed in a graph with a trend over time.
Avg. Helpful %
For each agent conversation, the % is calculated as thumbs up/total reactions, then applied, captured, stored, aggregated, and displayed here in a graph with a trend over time.
Tier zero support resolution
The percentage of total agent runs in which AI resolves the outcome without human escalation is captured, stored, aggregated, and displayed here in a graph with a trend over time.
Agent performance
The agent's overall performance is measured here:
- Agent
- Status
- Deflection rate: Deflections/ Runs per agent, as %,
- Helpful: thumbs up / total reactions
- Health:
- Green: deflection rate >= 80%,
- Amber: deflection rate b/w 70–79%
- Red: deflection rate <70%
- Run status: Active or inactive runs,
- Average completion
- Feedback: including thumbs up/thumbs down metrics
Channel usage
This is based on the channels the agent interacts with, such as Teams or Slack.
Most used tools
The most used tools by the agent, such as Knowledge base search, are displayed here with the number of use events and success rate.
Most effective topics
This is calculated by identifying the deflection rate. Topics with >=50% deflection are the most effective topics, and <50% are the least effective topics.
Least effective topics
This is based on thumbs-up/thumbs-down interactions and on applied feedback events captured, stored, categorized, and displayed here, along with conversations and percentages.