YUUKI EDGE — The Read · Issue #30 · October 6, 2026
Last week I wrote about AI note-taking tools and what happens to outside companies when the record vendor builds its own version. This week's news is the same question about a different tool, the one doctors use to look something up between patients.
On September 28, OpenEvidence confirmed a $250 million raise that values the company at $15 billion. It's a search tool where a doctor types a question in plain English and gets an answer with citations from medical journals. The company says more than half of U.S. physicians use it regularly.
The surprise is where it sits. Becker's reports OpenEvidence is built into Epic, the record system most hospitals run, at health systems including Mount Sinai and Sutter. Epic has also been adding content from UpToDate, a long-standing paid reference, to its own AI. So the same chart can now hold two rival ways to look something up, and the hospital pays for only one of them.
Three questions I'd ask about any tool doctors use at work
Who picked it, the doctor or the hospital? A tool a doctor chose on their own tends to spread quickly and quietly, with no purchasing process to slow it down. A tool the hospital bought arrives with the default setting on its side. Ask which one describes your product, or the product you're buying.
If the hospital changed the default tomorrow, would the doctor still open it? The real test is whether a doctor would pull out a phone to get to it. If the answer is "probably, yes," the tool has a habit behind it. If the answer is "only when it's already on the screen," it has a placement.
Who pays, and would they keep paying if it stopped being the default? Some of these tools are free to the doctor and paid for by advertising. Others are paid for by the hospital. Know which payer you depend on, and what that payer would do if a bundled version showed up.
How doctors look things up, and why the hospital's choice isn't always theirs
When a doctor isn't sure about a drug dose or a treatment guideline, there used to be one habit: open a paid reference such as UpToDate, which hospitals license for all their clinicians. Wolters Kluwer, which owns UpToDate, says it has more than 3 million clinician users in thousands of hospitals.
The newer tools answer questions directly. The doctor types it in and the tool writes a short answer with the journal citations underneath. Many of them are free to verified doctors. They spread because a doctor can start using one tonight, with no meeting and no contract.
Now the record vendor is putting licensed reference content inside its own AI, where it sits next to the patient's chart. That doesn't make the doctor's favorite tool disappear. It does give the hospital a way to make its own choice the default.
So here's the question I'd keep an eye on: when the hospital's record system offers an answer on the same screen as the one the doctor already prefers, which one does the doctor keep using?
The field splits into three groups, described by how they got in front of the doctor:
Tools the doctor adopted alone, usually free and paid for by advertising. Tools the hospital licenses and places inside the record. General-purpose AI assistants from the big AI labs, one of which launched a free version for verified doctors in April.
The named companies in each group, with calls on each, are in the paid issue.
Also this week EliseAI raised $350 million at a $4 billion valuation (MedCity News roundup) (announced September 29). The company builds software that answers calls and messages for apartment buildings and for specialty doctors' offices. Parakeet Health raised a $10 million Series A (announced October 1). It makes an AI assistant for the front desk of specialty practices, and says six of the ten largest dermatology groups use it. The FDA wants public comments by October 19 on how it should regulate AI tools built on large language models, the same technology behind chatbots. A quiet financing week otherwise. Those three rounds were the only healthcare-AI financings I could verify inside the window, and I'd rather say so than fill the list with older news. Where I might be wrong
Two studies published two weeks apart in June reached opposite conclusions about whether specialized medical tools beat general chatbots. One was peer-reviewed and the other was a preprint, and neither has settled it. If the evidence turns decisively one way, the winner changes. The three questions still hold, because they ask who chose the tool and who pays for it, and the answers to those don't depend on which tool is more accurate this month.
If you want the names
The paid issue, Yuuki Institutional, puts a name on every company in each of those three groups and scores each one on five things: how deeply it connects to the record, how good the evidence is, whether real customers have signed, whether doctors actually use it, and whether the record vendor can absorb it. Every score comes with a back, watch or pass call and the event that would change it. The founding rate is still open.
If you evaluate these companies for a living, the named calls, scores and what would change my mind on each one are in this week's Institutional issue: yuukiedge.com.
Got a minute? I'm asking the people who read Yuuki Edge what's making AI hard to get working in their organizations.
It's six short questions and takes about three minutes. I read every answer.
Link.
Yuuki Edge — the survival read on healthcare AI. Written by Victor Phillips, MD.