From AI Scribe to Clinical Copilot
Caroline Zhang, CEO of Knowtex, on going from AI medical scribe to AI lab
This is the move from note taking into full chart reasoning. Once an AI scribe can pull old labs, pathology, genetics, imaging and prior therapies into the same workflow, it stops being just a transcription tool and starts acting like a prep and follow through layer around the visit. In oncology especially, that matters because key facts are scattered across years of records, PDFs, scanned reports and specialty systems.1 2 3
- Knowtex already uses this data in concrete ways before and after the visit. It pulls abnormal lab trends, radiology impressions, pathology and genetic markers, then drafts summaries, visit agendas, staged orders and trial screening prompts for coordinators.1
- This is also where enterprise AI scribes actually differentiate. Basic note generation is getting bundled into EHRs and free physician tools, so the harder product is deep workflow completion, meaning the system not only writes the note but also fills diagnoses, orders and follow up tasks inside the record.1 3 4
- The catch is integration depth. Healthcare data is fragmented across many EHR formats and specialty systems, and generalized APIs usually do not go deep enough. That is why Knowtex uses official EHR integrations, computer use agents and forward deployed engineers to adapt workflows at each customer.1 4
The category is heading toward multimodal clinical copilots that sit on top of the record and do more of the real job. The winners are likely to be the companies that can turn messy historical data into usable actions at the point of care, while fitting into hospital systems deeply enough to survive bundling pressure from Epic, Oracle and other incumbents.1 2 3 4