AI company Knowtex has launched a dedicated frontier lab for healthcare, formalizing its work to build and rigorously test clinical AI tools that prioritize accuracy, auditability, and clinician trust. The move sharpens the company's focus on a sector where general-purpose AI often falls short, particularly in specialties demanding precise coding and longitudinal patient history.
From medical scribes to specialty-specific models
Knowtex was founded in 2022 by Stanford AI scientists Caroline Zhang and Jocelyn Kang. Before writing a line of production code, the founders embedded themselves inside clinical workflows. "We spent a combined year in scrubs as medical scribes, working forward-deployed alongside clinicians during the day and building the product at night," Zhang, the company's CEO, said in a statement. "That remains how Knowtex operates."
The company initially concentrated on oncology, a field where exactness in staging and coding is non-negotiable. Its platform now spans more than 200 specialties at over 300 organizations nationwide. Zhang said the lab launch commits Knowtex to publishing its models, AI harnesses, and evaluation benchmarks publicly.
Knowtex builds proprietary models for each specialty and layers on what Zhang calls "anti-hallucination safeguards" to protect record integrity. The system also includes embedded evaluators designed to judge clinical AI the way a practicing clinician would. This approach to evaluation science underpins a benchmark the company calls KnowBench, which measures effort reduction on administrative tasks. Knowtex's platform scores 97.99% on that benchmark.
"The clinician's role becomes review and signature rather than authorship," Zhang said. "Errors are built into the measure, because anything the system gets wrong has to be fixed by the clinician and registers as effort. Knowtex is publishing the full methodology so any system can be measured the same way."
VA contract and clinician adoption
In October, the Department of Veterans Affairs awarded Knowtex a $15 million contract to supply cloud-based ambient scribe tools that transcribe clinical encounters and generate notes. The technology is now live across 79 VA medical centers for more than 7,000 clinicians.
Zhang reported that VA clinicians have saved over 450,000 hours of documentation time, with an average satisfaction rating of 4.5 out of 5 across more than 5,100 ratings. Sustained clinician adoption sits at 88%. "Most tools in this category struggle to reach a third of users," she said.
A 90-day VA evaluation at the Kansas City site involving 18 primary care providers found that all participants wanted to continue using the ambient scribe after the trial. Most reported saving one to two hours of after-hours work. Patient experience scores at the site rose nearly three percentage points to 95.8% over the same period. One provider said the tool allowed "more time talking to patients, more eye contact, more time to use other resources while patients are talking."
The underlying Speech-To-Text technology that powers ambient scribes like Knowtex's is becoming a critical infrastructure layer in clinical settings, turning spoken encounters directly into structured notes.
Growth trajectory and product philosophy
Knowtex is cash-flow positive and has seen 10X revenue growth and 100X customer growth since the start of the year. Zhang framed the company's roadmap as an expansion beyond documentation.
"Documentation is where Knowtex started, but our mission has always been to maximize intelligence per patient encounter," she said. "Which means that when a clinician walks into a room, the full weight of that patient's history, the relevant evidence, and the right next steps are already working on their behalf, and when they walk out, the encounter has been captured completely and correctly so that the next clinician inherits that same intelligence."
Why this matters for healthcare professionals
Knowtex's lab launch signals a push toward verifiable, specialty-grade AI for Healthcare rather than general-purpose tools retrofitted for clinical use. For clinicians and health system leaders, the emphasis on published benchmarks and auditability offers a concrete way to compare vendors. The VA deployment data - 88% sustained adoption and measurable reductions in after-hours documentation - provides one of the larger real-world reference points for ambient scribe technology in a major health system. Knowtex plans to release additional public, clinician-grounded evaluation benchmarks, which could raise the evidence standard buyers expect from all clinical AI vendors.
Your membership also unlocks:
