The challenge
A busy physiotherapy clinic with 4 therapists was handling all appointment requests manually — phone, email, and WhatsApp. The front desk was fielding 40+ contact attempts per day, half of which came outside office hours. Missed calls meant lost bookings. Staff were spending more time on admin than patient care.
What we built
We built an AI scheduling assistant powered by Claude that sits in front of their existing Google Calendar. Patients interact via a simple web widget or WhatsApp number. The assistant checks real-time availability, books or reschedules appointments, sends confirmation and 24-hour reminder SMS via Twilio, and logs everything to Notion for the clinic's records. After-hours requests are handled exactly the same as daytime ones.
The outcome
Within two weeks of going live, 94% of all appointment bookings were handled end-to-end without staff involvement. The front desk now focuses exclusively on patients who are physically in the clinic. No-show rates dropped 28% once automated reminders were running.
The before state
Every booking required a human in the loop. A patient would call, get voicemail, leave a message, wait for a callback, and go back and forth before landing on a time. Outside of 9–5, requests piled up and were dealt with in batches the next morning — meaning patients often went elsewhere.
What we built
The core is an n8n workflow that bridges the AI conversation layer (Claude) with Google Calendar's API. Claude manages the conversational back-and-forth naturally — understanding intent like 'I need something Tuesday morning that's not too early' — and translates that into slot queries and calendar writes. Twilio handles all SMS. The whole system runs on the clinic's existing infrastructure.
How it runs today
The clinic owner checks a Notion dashboard each morning. All bookings, cancellations, and no-shows are logged automatically. Edge cases — like a patient wanting to book for someone else or requesting a specific therapist with a complex constraint — escalate to a staff member via a Slack message. These account for less than 6% of interactions.