Andy AI

Recording starts with two taps and runs in the background while the nurse examines the patient. The app handles noisy rooms and holds the audio locally when there's no signal, then uploads once the phone is back online. That matters in home health, where the visit happens in a living room, not a quiet exam room.

Andy AI drafts OASIS responses rather than handing back a transcript. It follows item intent and skip logic, produces the scored responses, and leaves each one for the clinician to confirm. The company ships updates when CMS changes the data set, which is the difference between a general scribe and a home health AI scribe. OASIS-E2 went live in April 2026, and older cheat sheets already miss items like A0810 and A1110.

Before the chart leaves the app, Andy AI cross-checks the visit against the patient's history and physical and their prior visits. It flags things like a hip replacement code with no surgical incision documented, or post-op pain that never made it into the narrative. Reviewers get a short list instead of a full re-read, which is where most QA hours disappear.

Coding is generated by the model and then reviewed by home health coders, with every code verified before it lands. The agency gets primary and secondary diagnoses mapped to the chart. Andy AI quotes 7% higher case-mix weight across agencies using the full stack, a figure the company measures itself, so treat it as a vendor claim rather than an independent benchmark.

The completed chart goes back into the agency's existing EMR instead of a separate portal. Andy AI says it works with the major home health EMRs, and setup happens during onboarding so patient history and prior visits load before the first recording.

Clinicians who work per diem can open a patient's prior visits and summary before walking in. That turns a blind visit into a prepared one, which matters for PRN staff who may not have seen the patient before.