Built a voice agent that records what payers actually say on denial calls — intro + looking for feedback
Hey everyone — just joined through the AssemblyAI Voice Agent Hackathon and wanted to introduce myself and share what I built.
I’m Chris Velez, a denial management analyst with 11+ years in healthcare revenue cycle. I’ve spent a lot of time on payer calls — the kind where a rep tells you one thing verbally, then the EOB says something completely different, and you have no record of what was actually said.
So for the hackathon I built The Witness — a voice agent that sits on payer calls and captures what’s spoken in real time. It detects contradictions between what the rep says and what’s on file, logs commitments and hold times, and generates a structured record you can actually use in an appeal. The core idea: evidence, not adjudication. The agent doesn’t interpret — it documents.
Built on AssemblyAI’s Voice Agent API with a Next.js frontend. Still in active build.
Happy to connect with anyone working in healthtech, RCM, or voice AI — and open to honest feedback on the concept. What’s missing? What would make you actually use this?
1 comment
Welcome to Lounge, Chris. Eleven years of sitting on payer calls shows in the problem you picked, since a verbal promise that vanishes by the time the EOB arrives costs real money. Who feels that pain most, the billing teams or the practice owners, and who would pay for it?