Knowtex's Clinical Decision Moat
Caroline Zhang, CEO of Knowtex, on going from AI medical scribe to AI lab
The real moat is not the model, it is the stream of doctor approved clinical decisions flowing through the product. Knowtex is trying to capture a richer training set than a generic scribe or medical chatbot can see, by combining chart context, the live patient conversation, and the clinician’s signed off plan. That lets it fine tune specialty specific systems and keep swapping in better base models underneath the same clinical workflow layer.1 2 3 4
- Most AI scribes only turn speech into a note. Knowtex is built around chart prep before the visit, then note, coding, orders, care coordination, and trial matching after the visit. That wider workflow creates more labeled examples of what doctors actually do, not just what they said out loud.1 2
- This matters because note writing alone is getting cheaper fast. Doximity launched a free Scribe in July 2025, and Oracle added order drafting to Clinical AI Agent in February 2026. In that market, vendors win by owning the clinical workflow and the feedback data, not by relying on a single foundation model.3 4 5 6
- Knowtex is positioning that data asset as enterprise grade from day one. It says the dataset is expert labeled, rights defined with customers and partners, and generated inside real clinical systems. That is also part of why integration strength mattered in its VA win, where workflow fit counted alongside model accuracy.1 2 7 8
Over the next few years, the category is likely to split between free note takers and deeply embedded clinical copilots. If Knowtex keeps converting each visit into reusable training data for specialty workflows, it can move from saving doctors time to shaping orders, trial enrollment, and treatment decisions at the point of care.1 2 3 4 7