What does scoping a first pilot look like for a healthcare administration team?

An environment team builds a simulator for hospital administration agents. Their first workflow is a procedure authorisation granted by a clinical reviewer with conditions: a specialist referral on file, and a pre-procedure assessment completed before the scheduled date. Cases use identifiers such as `AUTH-0412` and `CASE-0412`.

The team writes a pilot request covering:

- **Roles:** clinical reviewer, bookings officer, patient liaison and a billing clerk, with clearing authority marked for each condition kind - **Gated action:** theatre booking `BKG-0412`, blocked by both conditions - **Exceptions:** an assessment completed but dated after the booking, and a referral that expires before the procedure date - **Format:** JSONL per record type with a schema file, plus a snapshot for each task - **Checks:** foreign keys, status transitions, policy fit, and a clinician's read of plausibility

They leave volume open and ask for a small sample first. After inspecting it, they find that the billing clerk role never clears a condition, which suggests the authority model is too narrow, and they request a revised sample with an added waiver path.

**Consequence:** the wider request is scoped from evidence instead of guesswork.

**What it teaches:** keep the first request small, write the checks before the data arrives, and treat the first sample as a test of fit.