WhatsApp Booking Bot for Clinics

The conversation design phase was where this project lived or died. A bot that feels robotic drives patients away; one that feels too casual undermines clinical trust. I spent two days mapping every possible conversation path and crafting responses that were professional but warm.

The state machine architecture handles the core flow: location → service → date → time → confirm. But the real complexity is in edge cases: what happens when all slots are full? When a patient asks a medical question? When they type 'cancel' mid-booking? Each edge case got its own handler with a graceful fallback to human escalation.

Redis-backed session management means the bot remembers where each patient is in the flow, even if they come back hours later. 'Hi, you were booking a dental checkup at our Marina clinic — would you like to continue?' That continuity drives completion rates above 85%.

The analytics dashboard shows booking patterns by hour, day, and service type — giving the clinic data they never had from phone-based booking. They've since adjusted staffing based on peak booking times the bot surfaced.