How Marina became part of the Santamarina team, without ever being hired.

Context

Santamarina Clinic is a dermatology practice in São Paulo with a heavy patient flow: private-pay and insurance, first visits and follow-ups, questions about aesthetic and clinical procedures. A WhatsApp that receives messages around the clock, and a front desk that needs to attend to the patient standing right in front of it.

Every clinic that grows hits the same limit. Sunday night, Monday morning, any long holiday weekend, and the WhatsApp inbox piles up. Eighty, a hundred, a hundred and twenty unanswered messages. Each one is a person who may have given up, gone to a competitor, or who's going to show up irritated at the next interaction.

Santamarina's front desk couldn't keep up with everything at once. Attending in person and answering WhatsApp at the same time is a task nobody does well, and quality dropped on both sides. The in-person patient waited, the WhatsApp patient waited even longer.

The challenge

A generic chatbot wasn't enough. Dermatology has particularities that a menu-driven bot doesn't solve: a patient asking about an aesthetic procedure needs guidance that goes beyond scheduling. An insurance patient has different rules than a private-pay one. Someone who needs a follow-up has a different flow than someone booking a first visit.

The AI stepping in there needed to sound like someone from the clinic talking, not a system. It needed to know the procedures, the doctors, the prices (when applicable), the differences between insurance plans. And it needed to know when to hand off to a human, because some situations call for a sensitivity no machine has.

The solution

NexUnio built Marina, a virtual assistant for Santamarina, connected to the clinic's schedule, medical records, and management system. It's not a generic chatbot. It's a specific AI agent, trained on Santamarina's vocabulary, tone of voice, and protocols.

Today, Marina does what a good receptionist would do, and a bit more: answers every message at any hour; identifies the type of contact (first visit, follow-up, exam, administrative question, clinical question, rescheduling); checks the schedule in real time and offers compatible time slots; confirms the appointment directly in the system; sends reminders 48 hours before the appointment; escalates to a human when the case calls for it; and receives and replies to voice messages, because a significant share of patients prefer to send audio.

The integration with the clinic's system is what takes Marina out of the "chatbot" category and into the "team member" category. She doesn't simulate service: she performs it, with real persistence in the system.

The results

• Over 121,000 automated messages since going into production. That's 121,000 interactions that didn't consume front-desk time.
• Patient attendance jumped from 54% to 73%. Almost 20 percentage points. Automated reminders work, but they work even better when they come in the clinic's own voice, not a robotic SMS.
• 99% of patients don't realize they're talking to an AI. The measurement was done by sampling. Several patients praise the "receptionist" without suspecting a thing.

What this unlocks

Schedule freedom here means Santamarina stopped being held hostage by business hours. The patient researching at 11pm on a Wednesday now finds an answer. The person who walks into the clinic Monday at 8am no longer finds a backlog of accumulated messages, because they were handled over the weekend.

And the human front desk went back to doing what a human front desk does best: looking the patient in front of them in the eye.