Knowtex opened a clinical AI research arm on September 15, betting that proof of reliability will decide which vendors hospitals keep.
The company sells software that reads a patient visit and returns notes, billing codes, orders and after-visit summaries. It says the product runs at the U.S. Department of Veterans Affairs and at more than 300 other organizations.
The differentiator Knowtex is pushing is measurement. A new benchmark called KnowBench scores how much administrative burden a tool strips out of clinician work. The company reports a 97.99% result for its own platform. Hallucination controls and internal evaluation methods round out the lab’s work.
Caroline Zhang, the chief executive and a co-founder, argued that the real frontier is not model size. It is whether software can be trusted inside a live clinical workflow. Zhang and co-founder Jocelyn Kang met as Stanford freshmen studying AI and linguistics. They started the company in 2022 and each worked as a medical scribe.
Backing has followed. Knowtex’s investors include Jeff Dean, who until recently served as chief scientist at Google DeepMind. The company also reports that revenue grew tenfold during 2026 and that it is already turning a profit.
Documentation burden is a leading cause of clinician burnout, and U.S. healthcare administrative spending runs near $1T a year. Knowtex is arguing that reliability data, not raw capability, wins clinical procurement.
