Analyze AI agent's performance

Treat them as humans, check their KPI, and train them on better data to achieve best results

Written by Konstantine

Updated 2 months ago

The AI Agent Performance Report gives you a clear understanding of how your AI-powered chatbot agents are performing within a selected timeframe. This includes a breakdown of deflected vs. routed conversations, resolution metrics, and time-saving insights for the last 90 days.

How to access AI agent's performance report

Go to your HelpCrunch dashboard.
Navigate to Chatbot→AI Agents→Analytics→Performance.
In the top-right corner, choose a time period.
Select a specific AI agent from the drop-down list.

At the top of the report, you’ll find a high-level summary of:

⚫️ AI conversations – the total number of conversations handled by selected AI agent.
🟣 Deflected – conversations that were fully resolved by the AI agent without transferring to a human agent.
🔵 Routed to team – conversations that were escalated to a support rep.

In this sample timeframe, there were 383 conversations handled by AI agents. Out of those, 173 were deflected (45.17%) and 210 were routed to the team (54.83%).

Conversation outcomes

If you click the Deflected tab, this section breaks down what happened with the conversations that were not routed to the team.

🟢 Resolved – the customer’s question was successfully handled and marked as resolved by AI.
🔴 Abandoned – the customer left the conversation without resolving the issue.

💡 Low resolution rates can indicate that your AI agent needs more quality data sources.

Time-based metrics

If you scroll down the report, you will discover important time-related KPIs:

FRT (First Response Time) – Median time AI took to send the first response in a conversation. In the context of AI agents, value should be quite low.
AHT (Average Handling Time) – Weighted mean time AI interacted with a customer, regardless of whether the conversation was resolved, routed to a human, or abandoned.
Total Time Saved - Estimated overall time saved by AI in managing conversations, calculated using the Time to Close (TTC) metric for team members.

⌛️ In our example, the average handling time is 7 minutes 34 seconds, resulting in a total time saved of 3 hours 13 minutes for the team.

Summary

By regularly reviewing the AI Agent Performance Report, you’ll identify opportunities to improve automation, train your agents better, and reduce support load while keeping customer satisfaction high 🚀

Did this answer your question?