A health insurance agency’s journey from pilot to full-time voice AI agent

Managing a surge in customer inquiries during Medicare's annual open enrollment period can be challenging for health insurance providers. For eHealth, Inc., a US-based health insurance agency, handling the growing volume of calls became a significant operational concern. To improve accessibility and reduce customer waiting times, the company adopted an AI-powered voice agent called Alice, developed by Regal, to conduct preliminary screening for Medicare beneficiaries contacting its call center.

According to Atul Kumar, eHealth's Vice President of Product and AI, the Medicare enrollment window lasts approximately 54 days. During this period, the company handles around 500,000 calls while operating for 14 to 16 hours daily, including weekends. Despite these extended working hours, the demand for assistance can exceed the available capacity.

Although customers can enroll through eHealth's website, approximately 70% to 80% prefer purchasing insurance plans over the phone. This preference previously resulted in waiting times of up to six hours for customers seeking assistance from licensed human representatives. The company therefore needed a solution that could respond quickly while maintaining a professional, considerate, and customer-focused approach.

The deployment of Alice has helped address this challenge by answering incoming calls immediately. Data shared by Regal with CX Dive indicates that 77% of callers rated their experience as exceptional, while the purchase rate increased by 27% compared with the company's previous approach, which relied on outsourced human agents for initial screening.

From an After-Hours Pilot to Continuous AI Support

eHealth began testing voice AI in February 2025 through a pilot program designed to manage calls received outside regular call center operating hours. The initial implementation provided valuable insights into customer interactions, system performance, and the practical requirements of automated call handling. These findings helped the company develop Alice into a full-time voice AI screening solution.

Building a reliable agent required ongoing refinement rather than a straightforward deployment. Medicare enrollment involves personal circumstances, complex decisions, and conversations that can be emotionally sensitive. The AI system needed to recognize the context behind callers' statements and respond appropriately, even when people did not communicate their needs in conventional or predictable language.

Kumar described an instance in which a caller indirectly communicated that her husband had passed away. Alice interpreted the meaning of her words and responded with empathy. This example demonstrated the potential of large language model (LLM)-based voice systems to understand conversational context instead of relying exclusively on predefined phrases and keyword recognition.

Unlike traditional automated systems that often require extensive rules for anticipated responses, LLM-powered agents can interpret a wider range of natural language expressions. When implemented carefully, this capability can help businesses deliver more considerate interactions, particularly when customers need patience and understanding.

Adapting Voice AI to Healthcare and Insurance Regulations

Deploying AI in a regulated industry introduces additional technical and compliance requirements. eHealth must ensure that its customer interactions follow applicable healthcare and insurance rules, including requirements surrounding privacy disclosures and the handling of enrollment-related conversations.

One challenge involved delivering a mandatory privacy statement without interruptions. If a caller interrupted the statement, the process could require the disclosure to be repeated, creating frustration for customers and additional work for the call center.

To address this issue, eHealth collaborated with Regal to introduce a predefined, non-LLM-driven interaction step. This allowed the system to deliver the required statement consistently without the usual conversational interruptions. After the disclosure, Alice checked whether the caller was still available and then transferred the interaction to a human representative, regardless of the response.

The system also had to account for situations in which a caller was contacting eHealth on behalf of another person, including cases involving power of attorney. These scenarios can involve multiple participants and require the AI agent to handle the conversation appropriately.

Developing these capabilities required iterative improvements, practical testing, and close attention to real-world call behavior. The implementation team used A/B testing, reviewed recorded test interactions, and evaluated conversational quality alongside quantitative performance metrics. This combination helped identify issues that conventional performance measurements alone might not reveal.

Building Trust Through Collaboration and Continuous Testing

Introducing voice AI also raised concerns among eHealth stakeholders about inaccurate responses, inappropriate language, and potential AI hallucinations. To manage these risks, the company involved compliance specialists and representatives from quality assurance, telephony engineering, customer experience management, and training throughout the development process.

Senior leaders were also given opportunities to review actual AI-customer interactions. This approach helped them assess how Alice communicated with callers and understand the system's role in the customer service workflow.

The involvement of employees familiar with the demands of the Medicare enrollment period proved valuable. Their practical experience helped the development team establish appropriate conversational expectations and refine the AI agent more effectively.

The project illustrates why deploying AI in regulated environments requires more than selecting a model and connecting it to a phone system. Businesses must account for compliance obligations, unexpected customer responses, complex conversational scenarios, testing procedures, and clear transitions between automated systems and human representatives.

What This Means for Businesses Adopting AI Agents?

eHealth's experience demonstrates how AI voice agents can support customer service operations when demand exceeds the capacity of traditional call centers. By automating initial screening and managing incoming calls, organizations can improve responsiveness while allowing human professionals to focus on tasks that require specialized knowledge and personalized assistance.

However, successful implementation depends on aligning the technology with business processes, customer expectations, and regulatory requirements. Continuous evaluation, cross-functional collaboration, and human oversight remain important for maintaining reliable and responsible AI-driven interactions.

As businesses explore conversational automation, customized AI agent development can help them address industry-specific workflows, improve operational efficiency, and deliver more accessible customer experiences. Osiz Technologies supports businesses in exploring AI development and intelligent AI agent solutions tailored to their operational goals.

Voice Of Osiz Technologies

eHealth’s successful adoption of voice AI demonstrates how intelligent automation can transform customer service in high-demand industries. By using an LLM-powered AI agent, the company improved call accessibility, achieved strong customer satisfaction ratings, and recorded a higher purchase rate. This implementation highlights the importance of combining natural language understanding with empathetic and context-aware interactions. It also shows that AI solutions in regulated sectors require careful testing, compliance oversight, and continuous refinement. Integrating AI agents into customer service workflows can help businesses manage growing demand while improving operational efficiency. However, human involvement remains essential for complex decisions and personalized assistance. At Osiz Technologies, we see customized AI agent development as a valuable opportunity for businesses to enhance customer experiences and achieve measurable business outcomes.

Source: HealthCareDive.com