When I scope clinical AI, I separate drafting from rule-based checks.

In Arogya, typed or dictated rounds become editable AI drafts with the original source preserved. Deterministic rules handle checks such as allergy conflicts and abnormal-lab flags. Clinicians review and approve AI content before it reaches the chart.

That separation matters when a model or prompt changes. I'd want the team to rerun the rule checks with the same inputs and inspect their results independently of the draft's wording.

For a client evaluating an AI scribe, I'd ask to see why a flag appears and how the clinician reviews it before approving the record.

Public case: https://kapilchauhan.netlify.app/case/arogya.html