AI Scribes Require Deep EHR Access
Diving deeper into
Caroline Zhang, CEO of Knowtex, on going from AI medical scribe to AI lab
for most startups, the EHR is a death knell
Analyzed 4 sources
Reviewing context
The real moat in AI scribing is not the note model, it is getting deeply into the EHR workflow without becoming trapped by any one vendor.1 2 3
- If a scribe cannot write back into the chart, pull prior history, and help with orders, codes, and other follow up work, doctors still have to swivel back to the EHR and do the job manually. That makes the product feel incomplete and hard to justify in real clinical use.2 3
- That is why EHR access can kill startups. Large health systems usually require integration, security review, and business associate agreements before rollout. Small practices can start with copy paste, but moving up market means dealing with a fragmented long tail of EHRs and specialty workflows.2 4
- The twist is that EHR vendors have distribution, but not startup shipping speed. Epic aligned closely with Abridge and helped it scale from about 8,000 to more than 60,000 clinicians, yet that partnership also shows the tradeoff, fast access to workflow in exchange for revenue share, equity, and less freedom to compete head on with the platform.1 2
From here, winners are likely to look less like standalone note takers and more like AI workflow layers that sit on top of the record. The companies that win will be the ones that can plug into many EHRs, go deep enough to finish the work after the visit, and then use that position to expand into coding, orders, trial matching, and broader clinical intelligence.1 2 3
Conversation has been deleted
Start new chat