The intake that turns a concern into a priced offer

A clinic's own price list lists every procedure at once; a visitor with acne, sun spots or thinning hair does not know which line is theirs, or whether their case even qualifies for the procedure they picked. This demo intake flow for a Warsaw aesthetic clinic walks three languages, a guided questionnaire and an AI-written offer with a real price, ending in one tap to WhatsApp.

A client picks Ukrainian, Polish or Russian, then walks name, gender, one of twelve tap-only concerns (or free text) and whether they visited before. Claude reads the concern and writes three to six follow-up questions, tap-only, always including one gentle screening question for contraindications. The answers go to a second model call, which returns one priced offer: a single procedure, or a doctor's consultation when screening needs one, ending in a booking sheet with a direct WhatsApp link.

Before an offer reaches the client, a deterministic check confirms every price in it exists verbatim in the clinic's own price list and that any discount math is correct to the złoty. A model that fails gets one retry with the exact reason; a second failure returns a fixed, already-correct consultation offer instead of a guessed price. Per-IP and daily limits cap how many model calls a visitor can trigger, and any API failure sends the owner a Telegram alert.