LunaBill Is Using AI Voice Agents to Chase Down Insurance Claims, and It Is Working

The Macro: Healthcare Billing Is a $4 Trillion Phone Tree Nightmare

I want you to picture something. A medical billing specialist sits down at her desk at 8 AM. She has 200 outstanding insurance claims to follow up on. For each claim, she calls the insurance company, navigates an automated phone system, waits on hold for anywhere from 10 to 45 minutes, reads off the claim number, and gets a status update. She does this all day. Every day. For years.

This is not an exaggeration. This is how healthcare billing actually works in the United States. The American Medical Association estimates that the average physician practice spends $68,000 per year on interactions with health insurers. Across the industry, the total administrative cost of healthcare billing is somewhere north of $800 billion. A significant chunk of that is literally people sitting on hold.

The AI voice agent market has exploded. Bland AI, Vapi, Retell, Synthflow, and a dozen others are building general-purpose voice AI platforms. But healthcare billing is a specific enough domain that a general-purpose tool does not cut it. You need agents that understand CPT codes, denial reasons, appeal deadlines, and the specific IVR menus of every major insurance company. You need agents that can navigate a 12-step phone tree and then have a natural conversation with a human representative at the end of it.

That is a genuinely hard problem. And the market for solving it is enormous.

The Micro: A Billing Company Owner, a Healthcare AI Pioneer, and a Mayo Clinic Researcher

The founding team at LunaBill is unusually well-matched to this specific problem. Suhail Parry is a two-time founder who previously operated a medical billing company serving over 200 healthcare clients. He has spent three years in the industry. He knows exactly how painful claim follow-up calls are because he employed the people making them.

David Day founded Skiesoft, which became Taiwan’s largest healthcare AI scribe with over 120,000 users. He was also a data scientist at UCSF. Yash Raj is a Mayo Clinic researcher who built AI consumer applications. This is a team where every person has direct healthcare experience, not just AI experience applied to healthcare.

They came through YC’s Fall 2025 batch and the numbers speak for themselves. $764K in contracted ARR with $428K in live revenue. Over 50,000 calls automated. And the number that matters most: 100% pilot-to-paying customer conversion rate. Every single customer who tried LunaBill paid for it. That almost never happens in enterprise software.

The product deploys AI voice agents that call insurance companies on behalf of billing teams. Each client can run up to 100 concurrent agents. The system was trained on 1.2 million call transcripts, which means it has heard every hold message, every phone tree variation, and every claims representative response pattern across every major insurer. After each call, it delivers a structured summary that integrates into the billing team’s existing workflow.

The metric that jumped out at me is the 10x increase in claims processed per biller in the first week. Think about what that means for a billing company’s economics. The same person who used to process 20 claims a day is now processing 200. That is not an incremental improvement. That is a structural change in how the business operates.

LunaBill’s website is built on Framer and the product details are minimal on the public-facing site, which tells me they are selling through direct sales and pilots rather than self-serve signups. That makes sense for a product where the average contract value is probably in the tens of thousands per year and the buyer is a billing company owner or a hospital CFO.

The Verdict

LunaBill has the best early traction numbers I have seen from a healthcare AI startup in a while. $764K contracted ARR from a four-person team is real revenue, not pilot commitments or letters of intent. The 100% pilot conversion rate means the product works and the ROI is obvious enough that nobody walks away after trying it.

The risk is scaling the voice agent beyond claim status calls. Right now, the core use case is straightforward: call the insurer, navigate the phone tree, get the claim status, report back. That is automatable because it is repetitive and rule-based. The harder calls are appeals, where you need to argue with a human about why a claim should be paid. If LunaBill can handle appeals, the value per customer goes up significantly. If it stays limited to status checks, competitors like Thoughtful AI and Infinitus Health will eventually close the gap.

In thirty days, I want to see the contracted ARR number keep climbing. In sixty days, I want to know if they have expanded beyond billing companies into direct hospital customers. In ninety days, the question is whether they have started handling appeals and prior authorizations, which are the higher-value calls that billing teams hate even more than status checks. LunaBill picked the perfect entry point into healthcare AI. The execution so far is close to flawless. Now the question is how far they can take it.