From Scribe to Clinical AI Lab
Caroline Zhang, CEO of Knowtex, on going from AI medical scribe to AI lab
The real moat in clinical AI is not raw model size, it is getting doctors to trust the system enough to let it shape live care decisions. Knowtex is building around that bottleneck by sitting inside the visit, pulling messy record data before the appointment, capturing the conversation during it, and drafting orders, codes, and follow up after it. That workflow creates doctor checked data and makes adoption itself part of model improvement.1 2 3 4
- Knowtex started from the insight that note writing was only one small piece of the job. The harder work was chart review, order entry, coding, trial matching, and follow up. That is why its product grew from scribing into a full clinician assistant rather than staying a single note taking feature.1 2
- This matches the wider market. Abridge, Ambience, and Freed have all expanded beyond ambient notes into coding, chart prep, and workflow tools, because basic scribing is getting commoditized by free products and EHR bundles. Trust now comes from fitting into real hospital workflow, not from having a chatbot that writes a note.3 5 6 7
- Clinical trust is also a data strategy. Knowtex argues its advantage comes from the full clinical reasoning record, meaning the chart context, the visit conversation, and the clinician decision in one loop. That is more useful than a static dataset because it captures what the doctor actually considered and signed off on in care delivery.1 2 4
The next phase of healthcare AI will be won by products that become part of the exam room and the health system stack, then turn that usage into better specialty models and decision support. If Knowtex keeps compounding trust at the point of care, it can move from saving documentation time into influencing treatment selection, trial matching, and research workflows.1 2 4