Call Centers in the Age of AI

Host: Welcome to Current Trends for Payers. Today’s episode kicks off Season 4 with a conversation about something that affects every health plan: the contact center. We’re not talking about the old-school, press-1-for-benefits kind of call center. We’re talking about how AI is transforming these service hubs into strategic engines for member experience.

Here to walk us through the evolution of the call center to the contact center, is Reinaldo Toro, Managing Director of Contact Centers. Reinaldo has an interesting perspective because he was previously on the payer side, working with the vendor to implement a member engagement solution. Fast-forward to today, he is the leader for the digital contact center model, working with clients to oversee contact center operations. Reinaldo, welcome.

Reinaldo: Thanks for having me. It’s an exciting time for the industry. If you’re intellectually curious, you’ll enjoy witnessing the transformation from a call center to a contact center to a member experience center.  

Host: That’s an interesting statement. What do you mean by that? Call center to contact center to member experience center.  

Reinaldo: Call centers used to be about just providing information; get your answer, hang up. But over time, it became about how plans deliver that answer. We started realizing that this isn’t just a Q&A. This is a core part of the member experience.

Now throw in digital channels: secure email, live chat, member portals, provider portals, and suddenly you’re not running a call center. You’re running a contact center. And you’ve got to meet members on their preferred channel. Some want to chat. Some want to call. Some just want to get it done in the portal. And our job is to support all of it.

Host: So, how do plans operationalize and scale these new-age contact centers?

Reinaldo: The first thing to note is that all member channels must be able to offer the same information, which means providing real-time or near-real-time updates across all business units. You don’t want a member getting different answers from the call agent as compared to the member portal. There has to be consistency across channels. But, this is only possible with a consolidated data strategy where all the plan’s data is structured, connected, and flows across all areas of the business. This technology ecosystem is the foundation for the member engagement solution. Now, all member channels are informed by the same data. AI is only as effective as the data that informs it. That’s step one. With this foundation, AI steps in to support our call agents.  

Host: Can you provide some examples of how AI supports human agents?  

Reinaldo: Yes, there’s different types of support. Agent assist, virtual agents, chatbots.

Agent Assist gives information directly to the rep for support. In the future, we plan to implement and train it to access desk-level procedures and SOPs. So if someone asks about physical therapy benefits, the AI will know who the member is, what plan they’re on, and give the rep the exact answer in seconds. This is why the data must be connected through a whole-plan technology ecosystem.  

During the call, there’s real-time transcription. The AI listens as the rep and member talk, and it transcribes everything live. Reps don’t have to ask for a phone number twice, it’s there in the live transcription. And at the end, there’s a summary of the call. Reps can copy and paste it into their notes. This saves an enormous amount of time. As another example, if the member says they’re moving, the AI recognizes ‘address change’ and pushes a checklist to the rep: here are the five steps to follow, in order. They don’t have to remember it. They just follow the prompts.  

Host: So, let me make sure I’ve got this right. Agent Assist is AI working with the representative. It listens to the call, and it provides relevant info, or workflow suggestions directly to the rep?

Reinaldo: Exactly. It’s about enabling the agent to optimize the conversation. The agent can operate more effectively with workflow suggestions and consistency in how an issue is handled across different callers.

Host: Ok, so what is a virtual agent?  

Reinaldo: Virtual agents talk directly to the member. It sounds like a person, using natural language processing and large language models.  

Host: Is that the voice menu, “Press 1 for Claims, Press 2 for Benefits, Press 3 for ID cards?  

Reinaldo: Good question. What you just described is IVR, Interactive Voice Response. That’s being phased out in favor of a virtual agent. So instead of listing a menu and then having the member press buttons or scream ‘1’ into the phone, the virtual agent simply says, ‘How can I help you today?’ And the member says, ’I want an ID card.’ It uses NLP, so it sounds like a real conversation, and the interaction is more natural.  

Host: What are the different use cases for a human agent versus a virtual agent?  

Reinaldo: The first goal with virtual agents is to route members to the information they are looking for, quickly and efficiently. Virtual agents are ideal for handling simple inquiries. But if the member's question turns out to be complex, there are guardrails or conditions that are set to route the call to a human agent. This applies to chatbots too, which are just text-based virtual agents.  

Host: What are some examples of guardrails?  

Reinaldo: We can set thresholds. For example, if someone asks three or more questions, that may indicate a complex call, and that may get routed to a live agent. If they say ‘I want to file a grievance,’ we can set a guardrail that immediately escalates to a human. If AI detects that the member is upset or frustrated, we can have the call sent to a human agent.

Host: Oh, that’s an interesting use of sentiment analysis. For our listeners who may not know, sentiment analysis is an AI feature that detects human emotion.  

Reinaldo: Yes, we also use it to assist the human agent. If the member sounds frustrated, AI can support the agent in that moment with real-time suggestions to improve the interaction. So the AI isn’t just giving an answer to a question. It’s coaching the human agent through the interaction.

AI sentiment analysis also provides the opportunity to measure soft skills of each individual agent on all calls they have taken, rather than a random sample of calls. Prior to AI, random samples were selected, listened to, and analyzed by an actual person. It was a time-consuming process. Today, AI is used to analyze the quality of all member and provider inquiries.

Host: I'd like to discuss an AI feature that really sparked my interest for this episode. I caught a brief mention of it in a CNBC interview with Kevin Adams, where he mentioned accent neutralization. Can you explain what this is? 

Reinaldo: Oh yeah, this is a fascinating development. Accent neutralization is the overlay of an accent or dialect that differs from the human call agent's. So if your call center is offshore but you want it to sound like an American or British accent, the AI can do that. It’s already happening in the industry.

Host: That is interesting. Are there any ethical concerns? How do members feel about this so far?

Reinaldo: Careful consideration is given prior to the use of accent neutralization. However, people have responded very positively to the technology because it removes language barriers that interfere with communication. It increases clarity and understanding, and overall, it has received very positive feedback.

Host: That makes sense that it would reduce friction and ease the interaction if a member has difficulty understanding different accents. What about language translation? The contact center offers that too, right?

Reinaldo: Yes. right now, if someone calls and doesn’t speak English, we conference in a live interpreter. Our data shows us that 90% of those calls are in Spanish. Our roadmap includes AI-driven real-time language translation. Imagine the AI answers in Spanish, as the agent speaks their own native language in real-time. That’s where the industry’s heading.

Host: We’ve discussed using AI for sentiment analysis. Isn’t it true that AI can be used to help detect possible fraud by identifying certain behaviors and phrases in a request? 

Reinaldo: Yes, when people use social engineering tactics to execute a fraudulent activity, they often try to create urgency, such as, “I need my Medicare number right now, or I can’t see my doctor!” AI can be trained to flag certain phrases and behaviors to alert the contact center agent. And then provide assistance on how to handle the call. In the future, voice biometrics will capture each member’s voiceprint as a next-level identity verification. So, if someone calls in impersonating a member, the interaction would be flagged for fraud review.

Host: My ears perked up when I heard that health risk assessments are transitioning to AI. What’s the opportunity there?

Reinaldo: Every Medicare plan attempts to have health risk assessments completed annually. But they can be long and complex. You can’t do them through IVR prompts; it just doesn’t work. So we’ve been doing them human-to-human, or on paper. But now, we’re exploring the use of a virtual agent to guide them through the survey. And members can complete it conversationally, with the AI. That’s going to be a game-changer.

Host: Reinaldo, this is an incredible topic that the industry will continue to follow as the technology unfolds. With features such as real-time coaching of agents, improved consistency, fraud detection, multilingual support, and even voice overlays, AI is quietly transforming the payer contact center into a strategic member engagement hub. Thank you for educating us on what’s next for contact centers and member experience.

Reinaldo: My pleasure. It is an exciting time for the industry.

Host: Thanks for listening to Current Trends for Payers. If you found this episode helpful, share it with a colleague, and stay tuned for our next episode.