Why Healthcare Documentation Needs Structure (Not Just Better Prompts)
Clinical documentation is still one of the biggest friction points in healthcare and is hard to get right at scale. In this episode, we explore a programmatic approach to text generation that combines the strengths of large language models with deterministic structure.
Lasse and Victor break down why free-form prompt engineering often fails in medical settings, and how enforcing structure through techniques like JSON schemas enables reliable formatting, reduces hallucinations, and preserves each clinician’s individual voice.
We discuss ambient documentation, why “good enough” notes still require heavy editing, and how splitting generation into smaller, structured components unlocks faster, cheaper, and more trustworthy outputs.
This is a deep dive into how engineers and clinicians are reshaping clinical documentation from the ground up; not by replacing judgment, but by making it scalable.
