Trapeze Built AI Voice Agents for Doctor's Offices and Already Has 140 Practices Using Them

The Macro: The Phone Is Still the Front Door of Healthcare

There is a strange disconnect in healthcare technology. Billions of dollars have been spent on electronic health records, telehealth platforms, patient portals, and clinical AI. But when you need to see a doctor, there is a very good chance your first interaction is calling a phone number and waiting on hold while a receptionist finishes with the patient in front of them.

The front desk of a medical practice is a bottleneck that nobody has really solved. Zocdoc took a swing at it by letting patients book online, but most independent practices do not use Zocdoc because it charges per booking and does not integrate well with their existing systems. Patient portals from Epic and Athenahealth have booking features, but they are clunky and most patients do not bother setting up an account.

The result is that phone calls remain the primary way patients interact with doctor’s offices. And those calls are expensive. A front desk employee costs $35,000 to $50,000 a year, handles maybe 40 to 60 calls per day, and still cannot answer the phone after hours. Missed calls mean missed appointments, which mean lost revenue. For a small practice, every missed call is money walking out the door.

The AI voice agent space has exploded in the last year. Bland AI, Retell, Vapi, and a dozen others provide the infrastructure to build voice agents. But healthcare is different from scheduling a restaurant reservation or handling a customer service inquiry. Healthcare calls involve protected health information, insurance verification, clinical intake questions, and integration with EHR systems that were built in the 2000s. A generic voice agent platform does not handle any of this out of the box.

The Micro: Vanta’s Founding Engineer Meets a Healthcare Operator

Trapeze was founded by Christopher Chen and Betty Chang. Christopher was the eighth employee and a founding engineer at Vanta, then tech lead at Medallion. He knows how to build products that handle compliance-sensitive data at scale. Betty was Chief of Staff at Capable Health, with prior experience at Roivant Health, Intel, and the United Nations. She knows how healthcare organizations actually operate. They are part of Y Combinator’s Spring 2025 batch.

The product is an AI voice agent built specifically for medical practices. When a patient calls, the Trapeze agent answers, handles scheduling, collects intake information, verifies insurance eligibility, and integrates with the practice’s existing EHR system. It works 24/7, which means patients can book at 9 PM on a Sunday instead of waiting until Monday morning.

What sets Trapeze apart from a generic voice agent is the healthcare-specific training. The agents understand medical terminology, insurance workflows, and the specific way patients talk about their health needs. They can handle multi-step conversations that involve checking provider availability, confirming insurance coverage, and collecting pre-visit information, all in a single call.

The traction numbers are compelling. Over 140 doctors onboarded in the past three months, serving more than one million patients. For a two-person team, that is aggressive growth. It suggests the product is solving a genuine and urgent problem for practices.

The company describes itself as “AI-native Zocdoc,” which is a useful shorthand. Zocdoc built a marketplace where patients find doctors. Trapeze is building the operational layer that sits inside the practice and handles the work that Zocdoc does not touch. The two could actually be complementary rather than competitive, though Zocdoc might not see it that way.

The Verdict

I think Trapeze is attacking the right problem with the right approach. The front desk of a medical practice is an obvious automation target: high volume, repetitive tasks, significant cost, and clear pain when it fails. The fact that they already have 140 practices tells me the product works and the sales motion is efficient.

The healthcare-specific focus is the right call. Bland AI and Retell are building horizontal voice agent infrastructure. Trapeze is building a vertical solution for healthcare. History shows that vertical wins in regulated industries because the compliance, integration, and domain knowledge requirements create natural barriers. A generic voice agent can answer phones. A healthcare voice agent can answer phones while handling HIPAA, insurance verification, and EHR integration. That is a different product.

The risks are familiar for healthcare startups. EHR integration is painful and fragmented. Each practice uses a different system, and the APIs range from decent to nonexistent. Scaling from 140 practices to 1,400 requires building and maintaining integrations across dozens of EHR platforms, which is unglamorous and expensive.

The second risk is competition from the EHR vendors themselves. If Epic or Athenahealth builds native AI voice capabilities into their platforms, practices would probably prefer to use the built-in option rather than adding another vendor. Epic in particular has been investing heavily in AI features.

Thirty days, I want to see the retention data. How many practices that sign up in month one are still using the product in month three? Sixty days, whether they can move upmarket from independent practices to multi-location groups, which is where the real revenue lives. Ninety days, the EHR integration story. How many systems do they support, and how deep do those integrations go? The early traction is real. The question is whether the operational complexity of healthcare will slow them down or whether it becomes their moat.