Heidi's compute-heavy SaaS margins
Heidi Health
Heidi is selling labor replacement, not near zero cost software, so margin depends as much on compute efficiency as on pricing. Every visit triggers transcription, model inference, and secure cloud processing before a note can be pasted into the EHR. That makes the business look more like data heavy workflow SaaS, where profit improves when a doctor runs more visits through the system and fixed platform overhead gets spread across higher usage.
-
The product workflow is compute intensive by design. Heidi listens to the consultation, turns speech into text, generates SOAP notes, supports live voice edits, and runs added features like summaries and coding suggestions, all on private compliant infrastructure. Each extra workflow step adds value, but it also adds processing cost.
-
Bottom up competitors show the same economic shape. Freed sells a low price self serve scribe to individual clinicians, while enterprise players like Abridge charge much more by bundling deeper EHR integration and billing workflows. In both cases, the core product still starts with converting messy clinical audio into structured output, which is a metered compute problem, not a pure seat license.
-
The margin ceiling is also capped by infrastructure vendors. AWS HealthScribe and similar cloud tools let startups buy ambient documentation capabilities by the API call, which lowers time to market but makes the underlying engine easier to commoditize. That pushes independent vendors to win on workflow depth and specialty fit, not just on note generation alone.
The next phase is a race to turn each processed conversation into more revenue than compute cost. The winners in AI medical scribes will be the companies that use the same transcript once for notes, pre charting, coding, orders, and patient communication, because every extra output raises gross profit without requiring another sales motion.