Agent analytics shows how employees use AI agents and how effectively those agents resolve employee queries. Use it to understand AI activity, employee feedback, agent performance, and the topics and capabilities that have the greatest impact.
The dashboard includes activity and performance metrics, together with views that help you understand where employees use AI and where the AI performs best or needs attention.
Filter the data
Use the date selector at the top of the dashboard to choose the period you want to analyze.
The dashboard compares the selected period with the previous equivalent period, helping you identify changes in AI usage and performance over time.
You can also use available filters to focus the data on specific audiences or other supported criteria.
AI conversations
Shows the number of distinct conversations employees had with an AI agent during the selected period.
Each new exchange between an employee and the AI counts as one conversation, regardless of how many messages were exchanged.
Use AI conversations to understand how often employees are engaging with AI agents.
AI messages sent
Shows the total number of messages sent by AI agents during the selected period.
Only agent-generated responses are counted, not employee messages.
Use AI messages sent to understand how actively AI is communicating with your workforce.
Average helpfulness
Shows the average helpfulness score employees gave their AI conversations during the selected period, expressed as a percentage.
Only conversations where an employee explicitly rated their experience are included.
Use Average helpfulness to understand how employees rate their AI experience.
Deflection rate
Shows the percentage of employee queries that the AI assistant resolves independently, without HR involvement.
Only routine, self-service matters are included. Sensitive topics, such as harassment or disciplinary queries, are always escalated to HR even when the knowledge base could answer them.
Use Deflection rate to understand how effectively AI is resolving employee queries without HR involvement.
Agent performance
The Agent performance table shows how individual AI agents are performing during the selected period.
It includes:
- Deflection rate - the percentage of queries the agent resolves independently.
- Helpfulness - the helpfulness score for the agent's conversations.
- Times used - how many times the agent was used.
- Average response time - the agent's average response time.
- Employee feedback - thumbs-up and thumbs-down feedback from employees.
Use this table to compare agents and identify opportunities to improve configuration, knowledge, or agent performance.
Topics with high deflections
Shows the HR topics where AI is performing best by resolving the highest proportion of employee queries without HR involvement.
Use this view to understand where your AI is most effective and where self-service is working well.
Topics needing attention
Shows the HR topics where AI is least effective at resolving queries independently.
These topics may indicate knowledge gaps, configuration issues, or areas where employees are more likely to need HR involvement.
Use this view to identify topics that may need further investigation or improvement.
Conversations by channel
Shows how employee AI conversations are distributed across available channels, such as Teams, Slack, Web, and Email.
Use Conversations by channel to understand where employees prefer to engage with the AI assistant.
Most used AI capabilities
Shows the AI agent capabilities used most frequently to answer employee queries, such as searching the knowledge base or retrieving information.
Use this view to understand which capabilities are most important to how AI serves your employees.
A high usage count indicates that a capability is frequently involved in answering employee queries.
Best practices
- Set the period first. Compare equivalent periods when investigating changes.
- Separate activity from effectiveness. High usage doesn't necessarily mean high performance.
- Look at related metrics together. Review helpfulness and deflection alongside conversation volume.
- Investigate weak topics. Use Topics needing attention to identify areas for knowledge or configuration improvements.
- Compare agents. Use Agent performance to identify differences in deflection, helpfulness, response time, usage, and feedback.
- Consider the channel. Differences in usage across channels can help explain how employees are choosing to access AI.