Why Voice AI, Not Chatbots, Is Healthcare’s Answer to Patient Billing Support Costs

When you think of AI in customer service, you probably think of chat. After all, who hasn’t used a chatbot to change a flight or return a pair of shoes with just a few taps? It’s fast, it’s easy, and it gets the job done.

We thought that too. Our early work with large language models (LLMs) focused on AI-powered chat tools for patient billing support. And the results were promising: call center agents were able to get to chats quicker, and resolve them more efficiently.

But when we dug deeper with Cedar clients, we discovered something surprising: the vast majority of patient billing inquiries still happen by phone despite meaningful investment in digital channels. In other words, voice support is where patients and revenue cycle teams need the most help—and where healthcare organizations can unlock the greatest cost savings.

Why patients pick up the phone

It’s no secret that traditional patient billing support is broken. Whether it’s the cost and churn of staffing an internal call center, or the inconsistent service quality of an offshore BPO, the current approach isn’t sustainable. Especially in a world where patients are the fastest-growing payer in a provider’s book of business.

As more financial responsibility shifts to patients, they’re left to navigate healthcare’s byzantine payment process. Billing and insurance communications are rife with jargon. A single episode of care can involve multiple stakeholders, each with their own policies and procedures. And the bills? They often leave patients on the hook for hundreds, even thousands of dollars.

Is it any wonder the first thing patients do is reach for the phone? They’re looking for that human connection—someone who can show empathy, understand context, and provide reassurance.

The human touch of voice AI

The real advantage of voice AI isn’t just operational, it’s conversational. It comes down to the fundamental difference between reading and hearing. Voice creates connection in ways text simply cannot, even when both deliver identical information.

For healthcare billing—often a source of stress and anxiety for patients—this distinction matters. Voice AI can:

• Convey empathy through tone modulation, pacing, and natural pauses—subtle but powerful cues that text lacks entirely.
  
• Adapt its speaking style based on patient distress signals, slowing down when explaining complex charges or showing more warmth when sensing frustration.
  
• Provide immediate verbal confirmation that builds confidence and trust (“I understand you’re concerned about this charge from your recent hospital stay”).
  
• Respond naturally to interruptions, allowing patients to interject with questions or concerns without losing context—mimicking the back-and-forth flow of human conversation that’s impossible in text-based exchanges.

These patient benefits translate directly to tangible business outcomes. Voice AI can fully automate many billing inquiries without human intervention, reducing operational costs while maintaining—or even improving—patient satisfaction scores.

Unlike traditional IVR systems with rigid menus and touch-tone navigation, today’s voice AI allows patients to speak naturally. They can ask questions in their own words and receive compassionate support—addressing exactly what patients have been picking up the phone for all along.

Making the most of voice agents

The promise of AI voice agents is compelling, but delivering on it takes more than just plugging in new technology. It requires meaningful adoption.

Too often, providers jump straight to implementation—only to find that the voice agent frustrates patients or drives calls back to humans. To understand what people actually want from voice AI, we tested real agents across different industries. We observed callers interacting with automated voice agents for train balance inquiries, refunds, and other everyday tasks.

Based on our research,¹ here are four principles to keep in mind when rolling out AI voice agents:

1. Choice matters: Not everyone will want to use the voice agent. Some will go to great lengths to avoid speaking to a human, while others won’t. The key is providing the right level of support wherever patients choose to engage.
   
2. Time is precious: People need direct, concise answers to their questions. We found two-to-three minutes to be the sweet spot. Patients need to feel like they’re saving time, with no obstacles when they want to speak to a human.
   
3. Steer, but don’t control: The AI voice agent should guide the conversation, but the caller should always be in control. The solution should allow patients to change topics, interrupt, ask questions, and escalate to live support at any time.
   
4. Start simple, scale smart: Today, people prefer using voice AI for simple information lookup, saving more complex questions for human agents. However, as the technology evolves, we expect usage to expand.

What’s next for AI voice support

So, were our AI-powered chat innovations for naught? Absolutely not. They laid the groundwork for everything we’re building today.

Through chat, we learned what kind of quality to expect from LLMs in real-world scenarios. We saw how agents interacted with AI-generated responses in real time—what worked, what didn’t, and where the technology could be most effective. And we used those learnings to build a strong technical foundation for something even more ambitious.

That “something” is Kora, our AI voice agent purpose-built for patient billing.

Where chat made billing support faster and more scalable, voice takes it a step further. Kora delivers the human touch patients are looking for, with the consistency and efficiency that revenue cycle teams need. By prioritizing voice—where patients are overwhelmingly choosing to engage today and where healthcare organizations can realize significant cost savings—we’re meeting both patient preferences and business imperatives head-on.

In short, chat was our starting point. Voice is where it all comes together—and it’s only the beginning.

Learn more about Kora here

Ben Kraus is Director, Content Marketing at Cedar