How a Healthcare Company Improved Virtual Agent Containment with AI-Driven Call Analysis

Summary:

• Containment Rate Improved from 35% to Over 60%

• Reduced the number of transfers from the IVA to human agents

• Real-Time Monitoring for Latency, Security, and Accuracy

• AI Agent Improved with Issue-Level Reporting and IVA Optimization

Challenges:

A national healthcare company had implemented an Intelligent Virtual Agent (IVA) to handle two common inbound call use cases: invoice amount and invoice status. However, after a full year in production, the IVA was only successfully handling 35% of calls—leaving 65% to be transferred to live human agents.

This low containment rate meant the company was paying twice per call—once for the AI agent, and again for the human agent—negating the intended cost savings. They lacked visibility into what was causing transfers, which calls could have been automated, and whether the IVA was accurately responding to inquiries.

Solution:

To turn things around, the company partnered with ChatrHub to diagnose and resolve the containment issues. ChatrHub applied their GenAI analysis tools to identify breakdowns, recommend changes, and support continuous improvement of the IVA.

Key improvements included:

• Full Call Analysis from IVA to Human Agent Transfer: 

  • ChatrHub analyzed all IVA-handled calls, mapping them to those that were transferred to a live agent. This helped pinpoint exactly which topics or interactions caused failures.
    

• Identification of Inappropriate Call Routing:

  • Many calls were routed to the IVA that should have gone to a human agent—such as requests to update medical records. ChatrHub provided routing insights that enabled better upfront call steering.
    

• IVA Accuracy and Gaps in Handling:

  • For use cases intended to be automated, the analysis revealed whether the IVA was providing incorrect responses or simply lacked the logic to handle key follow-up questions.
    

• Root-Cause Breakdown of Transfer Drivers:

  • For calls that did escalate, ChatrHub broke down what specific questions led to a handoff—and what human agents did differently to resolve them and applied those methods to the IVA.
    

• Delay Awareness and Policy Mismatch:

  • A major issue was discovered: payments made by customers wouldn’t post for 48 hours, yet the IVA had no awareness of this delay. Customers would call after paying, see no update, and get transferred to a human agent because the IVA failed to relay this communication. These calls made up over 28% of transfers. Updating the IVA to inform customers of the 2-day processing window immediately reduced those transfers.
    

• Monitoring IVR Routing and Queue Congestion:

  • The company’s IVR offered limited options, creating long hold times on non-automated lines. As a result, customers frequently selected the IVA even when it was inappropriate, increasing transfer volume. ChatrHub flagged this routing issue and recommended changes to the IVR, which the company implemented to reduce inappropriate calls routed to the IVA.

Results:

After implementing ChatrHub’s recommended changes and enhancements, the healthcare company achieved:

• Over 60% Containment Rate—Up from 35% pre-ChatrHub

• Reduced AI-to-Human Transfers on Improperly Routed Calls

• Lower Cost Per Call by Avoiding Double Payment for Human Support

• Real-Time Monitoring for Latency and Fraud Detection

• Improved IVA Responses and System Awareness of External Delays

Conclusion:

This company was spending heavily on AI automation without seeing the return—until they gained the right visibility and support. By identifying call routing errors, fixing IVA gaps, and monitoring key performance issues like latency and hallucinations, ChatrHub helped them double their containment rate and reduce unnecessary spend.

If you're exploring AI agents for your contact center, we strongly recommend starting with a comprehensive AI Readiness Assessment. Our system analyzes human-to-customer interactions to uncover high-impact automation opportunities, estimate per-topic costs, and determine the required integrations for successful deployment.

Post-implementation, we also help you monitor your AI experience in real time—including resolution accuracy, latency, hallucinations, fraud detection, and inadvertent data exposure—ensuring a safe, efficient rollout.

Thinking about deploying a virtual agent? Start with the right foundation.

BOOK A DEMO