How R1 left the IVR behind and built something better for patients.

This article was co-authored by Ana Maria Nicolau, SVP, Operations at R1 and Miranda Zhao, Agent Development Lead, Healthcare at Sierra.

For decades, the Interactive Voice Response (IVR) has been the universal constant of healthcare customer service. Press 1 for billing. Press 2 for scheduling. Miss an option, start over from the beginning.

It worked. But as patient expectations evolved and the complexity of healthcare administration grew, R1 started asking a harder question: what if the phone call itself could actually be a good experience?

R1 sits at the intersection of providers, payers, and patients, handling the revenue cycle — everything non-clinical — on behalf of health systems across the country. When a patient gets a bill and picks up the phone, they're often talking to R1, even if they think they're talking to their hospital. The experience they have in that moment matters, and R1 saw an opportunity to make it meaningfully better.

A new way to think about the patient call

The traditional IVR was built to route calls efficiently, which was a genuine feat of operational engineering for its time. But the world has changed. Patients today expect more from every service interaction, and the technology now exists to deliver it. R1 started exploring what it would look like to replace the phone tree with something genuinely conversational, and built an AI agent that could meet patients where they are, understand what they need, and get them there faster.

The challenge in healthcare is real. The patient population is diverse, and data lives across multiple systems, including different electronic medical records (EMRs), separate scheduling platforms, layered phone infrastructure, and more. Unlike a retailer who can see your recent purchase, R1 operates under strict HIPAA requirements that shape every design decision. Any solution had to work within those constraints while still feeling like a step forward for the patient.

R1 was confident it could be done. And they were right.

Replacing the IVR with an AI agent

R1 launched its Phare Revenue Operating System in October 2025, a platform designed to deliver real-time operational intelligence and workflow orchestration across the revenue cycle. An AI-enabled contact center plays a key role in achieving this.

R1 deployed an AI agent to handle the calls that make up the highest volume of patient inquiries: balance checks, payment processing, payment plan setup, account questions. The goal was never to automate everything. It was to right-size the work, making sure calls that required real human judgment and care received exactly that, while patients with straightforward needs could get answers quickly, without a wait.

The design of how the agent operates turned out to matter just as much as what it could do. The most important decision made was to let patients lead. Rather than moving through a prescribed flow, the agent opens by asking why the patient is calling, and follows from there. Patients who feel heard before they feel helped respond differently. Engagement went up. Calls became more fluid. The experience started to feel less like a system and more like a conversation.

R1 also invested heavily in the data infrastructure underneath the agent. Healthcare data is notoriously fragmented, and an AI agent is only as useful as the information it can access. R1 did the work of connecting the systems — multiple EMRs, scheduling platforms, account databases — so that the agent could give patients complete, accurate answers. A caller with accounts across multiple service lines can now get a single summed balance in seconds. Verified patient identity passes downstream across departments, so a transferred caller doesn't have to re-authenticate. These aren't small things. Anyone who has navigated a healthcare call center knows exactly how much they matter.

Lessons learned along the way

Like any meaningful innovation, this one came with a learning curve.

Authentication was one of the most instructive chapters. The initial design verified patients first, then addressed their needs. While logical, in practice this meant patients were waiting over a minute before the conversation could begin. When R1 reordered the experience, establishing intent first, then authenticating, everything improved. Auth time dropped by more than half. A key learning here was that verification works best when it's based on numeric data while ensuring compliance requirements are met. Small adjustments, significant improvement.

The R1 team also learned something important about the pace of trust. There were calls where the agent handled everything correctly (answered every question accurately, walked the patient through exactly what they needed) and at the end, the patient still asked to speak with a human. Not because anything went wrong, but because trust in AI is something that builds gradually, through repeated positive experiences. Every good interaction is a deposit. Over time, the balance grows.

What's made all of this possible is the ability to iterate quickly. In the old world, changing a call script meant retraining hundreds of support reps across multiple sites. Today, we can test, learn, and adjust almost in real time, which means continuously improving the experience in ways that simply weren't possible before.

The results

The results speak for themselves. The AI agent is already resolving up to 40% of incoming calls in early testing. A patient can call and describe what they need in plain language, like "my dad told me I need an x-ray", and get routed correctly, no menu required. A caller with questions about multiple accounts gets a clear, complete answer in seconds. Patients who need a human agent arrive at that conversation already informed, with context established and the handoff seamless. The agent can focus entirely on solving the problem.

The contact center is becoming what it always should have been: a unified, intelligent front door to the health system. Not a gauntlet to navigate, but a resource patients can actually rely on.

Advice from R1

The R1 team shared their most crucial learnings: Start with the use cases that are high-frequency and straightforward. Early wins build confidence for your team, your compliance stakeholders, and your patients. Invest in your data infrastructure before you invest in the experience layer; the two are inseparable. And design your handoffs to humans as carefully as you design everything else. A patient who reaches an agent quickly, with their context intact, walks away trusting the system.

Most importantly: measure what matters. Automation rates are a means, not an end. Patient experience is the end. When you optimize for that, the efficiency follows naturally.

Looking ahead

This is only the beginning. For R1, there are more use cases to unlock, more data to connect, more moments in the patient journey to reduce friction and increase clarity. The work to date reflects real momentum and sets a strong foundation for what comes next.

A phone call shouldn’t be frustrating. For healthcare patients calling into R1, it’s an increasingly seamless experience.