A first-layer web assistant
— the smallest credible build we can put in front of the hospital this round.
V2 narrows the prototype to an omnichannel AI assistant — web and WhatsApp on one engine, priced on prepaid credits, with no human in our reply path. If the hospital wants a human to take a conversation over, that lives in their CRM, not ours. Work primitives and reminders are deferred to v3 — but the assistant judges itself by the same clinical line, refusal floor and reviewability that v1 specified.
Hospital site
+ assistant widget
Faux RS Telogorejo homepage. The assistant lives as a floating bubble. All five PRD scenarios run end-to-end — information, symptom-routing-to-booking, emergency, low-confidence refusal, zero-credit hard stop.
Admin panel
for hospital staff
Conversations, knowledge base, credit ledger, analytics, and settings. Hospital staff curate and review on the back end — they never speak to patients through this surface.
V2 deck
for the pitch
The PRD as a presentation. Frames the scope-down, names the tensions with v1 honestly, walks the credit model and the architecture, lands on the open questions Renlester is bringing to the call.