Shifting Intelligence Out of EHR
Caroline Zhang, CEO of Knowtex, on going from AI medical scribe to AI lab
The real opportunity is shifting intelligence out of the EHR and into a separate workflow layer that tells clinicians what to read, write, order, and follow up on. In that model, the record stops being the place where work happens and becomes the system where finalized data gets stored. That is why companies like Knowtex push beyond note writing into chart prep, coding, staged orders, and trial matching, because the more tasks the overlay completes, the less the underlying EHR owns the user experience.1 2 3
- This only works if the overlay goes deep into workflow. A shallow tool that drafts a note and asks staff to copy and paste still leaves diagnoses, orders, flow sheets, and billing steps inside the EHR. The competitive line is not whether a tool connects to the record, but whether it finishes jobs two through five after the note.2 3
- The tradeoff is integration depth versus EHR independence. Abridge grew fast by partnering tightly with Epic and Athenahealth, which gave it distribution into large systems, but that also ties it to the platform owner. Knowtex is taking the opposite path, building official connectors and computer use agents so it can sit across different records and act more like an independent operating layer.1 2 3
- The market is pushing this way because ambient documentation is getting cheaper and more bundled. Free tools from Doximity and OpenEvidence are training doctors to expect basic note generation at no charge, while Oracle and Epic are adding their own AI features inside the record. That raises the bar for overlays to prove value through full workflow automation, not transcription alone.3 4
The next phase is a split market. EHR vendors will keep absorbing generic scribing, while independent overlays win where they can automate specialty specific work across prep, documentation, orders, coding, and follow up better than the record vendor can. If that happens, the EHR becomes infrastructure, and the AI layer becomes the product clinicians actually feel day to day.2 3 4