An intake assistant that books the right appointment

The clinic has five practitioners with different specialties, and patients rarely know which one they need. Every booking call started with the same three or four triage questions, and the front desk answered them roughly two hundred times a week.

The clinic had tried a generic booking widget. It failed for a specific reason: it let patients pick any practitioner, so a third of appointments landed on the wrong calendar and had to be moved by hand, which cost more time than taking the call.

We started by sitting with the receptionist and writing down the actual decision tree she uses. That document, not the technology, is what made this work.

The assistant asks the same questions in the same order, in French or English, and only offers the slots that match the answers. It never diagnoses and it never speculates: when a patient describes anything outside the tree, it hands the conversation to the front desk with the transcript attached, which happens on about one conversation in eight and is treated as a correct outcome rather than a failure.

It also handles the boring half of reception: confirmations, reschedules, parking and what to bring.

Roughly seventy percent of bookings now happen without a phone call, most of them outside opening hours. Misrouted appointments have effectively disappeared, because the assistant cannot offer a slot the answers do not support.

No-shows fell by about a third, which the clinic attributes to the automatic reminders more than the booking flow itself. The front desk did not shrink. It moved to the patients standing in front of it.