Notes from the Field

Last time I wrote about why healthcare admin needs automation that’s deterministic and auditable. Since then, I’ve been thinking about a gap in that argument: a system can explain every decision it makes and still not be trusted. In coincidental timing, OpenAI released their report on the Hugging Face incident last Thursday; it was quite a read. It’s an extreme case of agentic coordination (and a...Read more

My two cents on the ABA toolkit: It is what it claims to be: guidance. And while I welcome the flexibility it gives states, I do think it’s a missed opportunity for federal mandates that states could implement at scale today. If you’ve been paying attention to federal and state action, what’s listed around documentation, licensure, and UM won’t come as a surprise. We’ve seen firsthand how leavin...Read more

CMS released its long-awaited autism toolkit for states today, consolidating guidance on medical necessity, credentialing, utilization management, fraud detection, and other areas for Medicaid and CHIP programs. The first thing that stood out to me: none of this is groundbreaking. States have been moving in this direction for the last few years, and many that took early action are already doing m...Read more

Of late, I've been spending a fair amount of time evaluating different agents, and it's changed how I think about the role of AI agents in healthcare. The obvious use case (and I think where most of healthcare is focused) is to give agents a task and let them go do the work themselves (eg. Cowork). But, at least in healthcare, I believe the biggest value often comes from using agents to build the...Read more