You already know: enterprise RAG (retrieval, permissions, citations), the support agent (deterministic checks around a model) and legal analysis (evidence verified by code). Clinical Q&A combines all three, with a twist that changes every trade-off: a wrong answer can harm a patient.
How to read this lesson. No code, on purpose. Numbers marked assumed are assumptions you state out loud; prices are illustrative. This is a systems design exercise, not medical or regulatory advice: in a real project, clinical governance and regulatory specialists are part of the team from day one.
1. The prompt, and the questions that change the design
A staff candidate notices quickly that scale is not the problem (a hospital network asks tens of thousands of questions a day, not millions a second). The problems are grounding (every claim traceable to an approved, current source), safety (some questions, like dosing, must not be answered by free text generation at all), privacy (patient data is among the most protected data there is) and governance (who is accountable for what the assistant says).
Assumptions you state: