Almanac launches an AI teammate for physicians
Almanac's AI assistant plugs into a clinic's existing tools, covering evidence retrieval, ambient scribing, medical calculators and clinical trial search, with each conversation in its own sandbox and permission asked before it acts.
Physicians log nearly 2 desk hours for every hour with a patient. That is a 2016 Annals of Internal Medicine number, and it holds even though every physician I know runs a delegation chain. The front desk runs the schedule, a biller handles the claims and a scribe writes the note. What stays on the physician's plate is the clinical work itself, the literature check and the risk score. Almanac wants that part. The pitch is an AI teammate that slots into the same chain. Each conversation runs in its own compute sandbox, so a physician hands off a multi step task and comes back to it later, and along the way it asks clarifying questions and asks permission before it acts. The company says the tools are plug and play, with evidence retrieval, ambient scribing, medical calculators and clinical trial search on by default, plus whatever the clinic already runs, EHR included. Cyril Zakka is a physician researcher out of Stanford who wrote a NEJM AI paper on retrieval for clinical medicine, one the company describes as among the journal's most cited. That matters because retrieval anchored in peer reviewed evidence is the hard part here, and it is what a physician will check before trusting anything else the tool does. Delegation only works when the physician can check what comes back, which is why the permission step matters more to me than the tool list. A tool that pauses to ask before it acts builds that checkpoint into the workflow. Physicians are going to keep score in hours handed back, starting with the first task they let a tool finish on its own.
This post quotes another post: x.com/i/web/status/2102782626297675864