Heidi launches Heidi II, an agentic layer that does clinical follow-up work

Heidi Health said Tuesday it launched Heidi II, moving its clinical AI from visit notes into agentic workflows that can pre-chart rounds, chase pathology, fill referral forms, and run saved Routines while clinicians stay in control of what leaves the system.

Clinical AI that only drafts notes still dumps the follow-up pile on the doctor. Heidi II is the bet that an agent can finish the paperwork chain with a human gate before anything hits the chart or the patient.

On Tuesday, 29 September 2026, Heidi Health announced Heidi II. The company blog, bylined Heidi Team and dated September 29, 2026, calls it an agentic layer for the clinical work around a visit, not only the note. Agentic, here, means the software carries a task through instead of only writing text. Heidi says the first wave of clinical AI took the visit note off the list, and that software which summarizes a visit without doing the follow-up still leaves the paperwork on the doctor. The page does not print an hour. Those lines are Heidi’s.

Heidi II opens on the clinician’s day. The new Home view brings scheduled patients, open tasks, and clinical prep onto one screen, instead of an empty box waiting for a prompt. Heidi says that screen surfaces the work a clinician is ready to hand over. Yekutiel Sandman, MD, of Urology Specialty Care of Miami, is quoted on the page: the notes are prepared and the relevant information is already there, so the day runs smoother and charting takes less time. That quotation is his, on Heidi’s page.

Agents can carry clinical and administrative tasks through before, during, and after the visit. The examples Heidi gives: ask it to pre-chart tomorrow’s ward round and it reads the charts overnight, so the clinician walks in knowing what changed. Ask it to close out today’s visits and it chases outstanding pathology, fills the referral form from the letter that came with it, and drafts the handover letter to the nurse. Pathology, here, means lab results that are still outstanding. A referral form sends a patient to another clinician. A handover letter passes the patient to the next person caring for them. The clinician reviews what needs a clinical decision. Heidi says you ask in plain language, and it works the task through to the end across integrated systems or older portals, then brings back what needs that decision. Around ten million tasks surface in Heidi every week, the company says. That is about ten million pieces of follow-up work showing up in the product each week, Heidi’s count. Until this launch, Heidi says, it could only show which tasks needed doing. Now it handles them. Raymond Brown, MD, of Capital ENT & Sinus Center, is quoted: Heidi loads his schedule, finds the pathology results in the chart, and writes a pre-visit note he can read in thirty seconds. That quotation is his, on the page.

A saved task can become a Routine. It runs the same way every time, on a schedule the clinician sets, for that clinician or the whole practice. Heidi says nothing is skipped silently, and anything it could not finish comes back named with a reason. The page also says a clinician can turn on a ready-made Routine, built by clinicians who already run it every day.

Memory personalizes Heidi for how that clinician practices. Heidi says it learns unwritten rules, clinical standards, and formatting preferences from direct instructions and from every correction, so the clinician only has to show it once. A general model, Heidi says, hands everyone the average. These lines are Heidi’s.

Heidi II brings peer-reviewed research into the workflow. From inside a session, a clinician can ask a question, and Heidi says it answers from the full text of articles in the New England Journal of Medicine, Wiley, and Cochrane, and more, read alongside the patient’s history. Cochrane publishes reviews of medical evidence. Wiley, here, is the journal publisher Heidi names. Answers come with citations, and journal subscribers can follow those citations to the full article. Heidi says an abstract compresses a paper, and that the detail cut from an abstract is often the part that would have changed a decision. Those lines are Heidi’s.

The control model, as Heidi states it. Heidi starts in Manual mode, so a clinician can watch the work and approve it as it goes. By default, anything that touches a chart, reaches a patient, or leaves the system waits for review, and a run can be stopped at any point. A chart is the patient’s medical record. Once a clinician trusts how Heidi handles a task, that task can be switched to Auto mode. Some actions stay off limits in every mode. Heidi will not prescribe, delete a record, or submit a claim. Heidi says Heidi II ships under the same independently audited certifications it already holds. The page does not name those certifications.

The rollout, as the page states it. Heidi II begins rolling out globally on the day of the post, in English, with more languages in the coming weeks. The features arrive in an account automatically. On a managed account, Heidi says its team will work through the setup, and where a customer needs something that does not exist yet, Heidi will embed engineers to build it. Heidi II is coming to Asia and the Middle East, and is not available in the EU and the UK. The page does not print a price, and it does not announce new funding.

In plain terms, Heidi said on Tuesday that Heidi II can pre-chart a ward round, chase pathology, fill a referral form, draft a handover letter, and run saved Routines, while a clinician still reviews anything that touches a chart, a patient, or a message that leaves the system. It will not prescribe, delete a record, or submit a claim. The English rollout starts with this announcement. The EU and the UK are not included.

The picture is Heidi’s Heidi II product screen from the launch page. It shows a visit note in SOAP form — subjective, objective, assessment, and plan, the usual four-part clinical note — an Ask Heidi panel, and a strip of journal names that includes NEJM, Cochrane, Wiley, and BMJ. NEJM is the New England Journal of Medicine. BMJ is the British Medical Journal. The same screen shows task rows for chart work, including actions labeled for Epic, the electronic record many hospitals use, such as pulling a patient chart and scheduling a follow-up. The blog text describes chart and referral work. It does not announce an exclusive arrangement with Epic.