I keep a list of LLM-driven startups in healthcare. Most of what I track falls into two buckets: Operations (scheduling, billing, prior auth) and Clinical (scribes, summarization, care navigation). Both assume the user is someone inside the healthcare organization. Two ideas keep nagging at me that don’t fit either bucket – they sit on the patient’s side.
Charity care. Most hospitals have a formal policy that offers income-based discounts or outright bill forgiveness. It’s required. It’s documented. And it’s massively underutilized, because navigating it requires deciphering policy language, collecting paperwork, and making phone calls. I volunteered with DollarFor, a non-profit that helps patients do exactly this, so I’ve seen firsthand how real the impact is and how manual the work remains. This is textbook agent territory: read the policy, understand the patient’s situation, determine eligibility, walk them through the paperwork. The information asymmetry is enormous, and the cost of not solving it falls on the people least equipped to absorb it.
Personal health navigator. When I was researching mitral valve replacement for a family member in late 2024, the volume of medical literature, patient forums, and clinical guidelines I accumulated was overwhelming. So I uploaded over 2,000 pages into Google’s NotebookLM and started having conversations with it. I wrote about that experience in detail. The most counterintuitive finding: the more information I fed it, the better the conversation outcomes became. That’s the opposite of the old reality, where more information just meant more overwhelm. RAG-based tools that serve as a personal navigator for individuals facing a specific health decision feel inevitable to me. The question is form factor, not feasibility.
These two ideas have been rattling around for a while. Neither fits neatly into the categories I already track, but both feel worth noting. If you’re building in this space, I’d like to hear about it.