Four clinics shared one reception team and a phone that rang while a patient stood at the desk. An AI front desk now answers around the clock, booking against the real calendar.
Hugo takes the project apart on screen: the bottleneck we found, the system we shipped, the number it moved.
No walkthrough yet. The written case below carries the same numbers.
Marbella Dental Group runs four clinics on the Costa del Sol with 38 staff and eleven chairs. Patients pay privately, most of them come back, and a good share fly in from northern Europe.
Reception held the phone, the front desk and the WhatsApp queue at the same time. Two receptionists covered four sites on a rota. Whoever was standing with a patient stopped answering the phone.
Missed calls were counted for the first time during the audit: 214 in ten working days. Nineteen of those callers left a voicemail. The rest were never called back.
No-shows sat at 18 percent. Reminders went out by hand whenever somebody remembered. Monday appointments rarely got one at all.
Two days of interviews and one week of their own data. Here is what was actually going wrong, told the way the people living with it told us.
Reception was two people doing three jobs. Whoever stood at the desk with a patient could not also pick up the phone, so the phone lost.
The callers who gave up stayed invisible until we counted them. Most were new patients asking about a first appointment, the one call a private clinic cannot afford to miss.
Reminders were the other quiet leak. They went out when somebody remembered between patients, which on a busy Monday meant they did not go out. Empty chairs turned up two days later, and nobody joined the two facts.
Built inside the tools the company already paid for. Every workflow, credential and prompt was handed over at the end.
Text and voice on the same brain. It identifies the patient, answers the question it can answer, and books when the calendar allows. Anything clinical goes to a human immediately.
WhatsApp
Vapi
ElevenLabs
OpenAI
n8nEvery appointment gets a confirmation at seventy-two hours. A cancelled slot goes straight to the waiting list, offered by WhatsApp to the three nearest patients.
WhatsAppn8n
PostgreSQLElevenLabsOne shared inbox where the group sees every conversation, human or agent. Reception picks up mid thread with the full history in front of them.
ChatwootWhatsAppn8nPostgreSQLBefore figures come from the company's own records, pulled during the audit. The after column is the ninety day average once every system was live.
| Metric | Before | After |
|---|---|---|
| Calls answered | 61% | 99% |
| Enquiries closed without staff | 0% | 68% |
| No-show rate | 18% | 10.6% |
| Reception hours on the phone, daily | 6 h | 40 min |
| New patient insurance check | 22 min | 3 min |
| Bookings taken out of hours, weekly | 0 | 27 |
The agent takes the calls nobody could take. My receptionists are back at the desk with the patient in front of them, which is what we hired them for.
Every project has them. Writing them down is how the next one gets shorter.
We launched text and voice together. Text was ready in week three; voice needed another month of tuning, and it held the whole project back.
Starting with WhatsApp alone would have handed the clinics a win four weeks earlier.
We also let the agent quote prices before the group had agreed one price list across the four sites.
First system in production inside a month. Everything after that is iteration with the client in the room.
First response, was 4 h: 11 min
Read the caseTickets answered under 15 min: 74%
Read the caseThe AI and Operations Audit finds it, prices it, then hands you a plan in two working days. Your fee comes off the first thing you build with us.
