The patient asks, the agent books

The centre brings together more than eighty specialists under one roof. For the people who work there it is a strength, for the patient it starts out as a maze. They know something is wrong but not which specialist they need, or which test has to come before the visit. They want to know whether to come fasting, what they can eat the day before, how much they will spend, which days the doctor is in.

All of these questions arrived in the same place: the phone and the front desk. Staff heard them repeated from morning to evening. At peak hours the lines were busy, and people who could not get through to anyone risked putting off the booking.

On closer inspection, the answers already existed. They were scattered across price lists, specialists’ timetables, test preparation instructions and internal procedures. What was missing was a way to give them straight away, at any hour, and to take the patient all the way to the appointment without tying up a person.

• By chat and by voice, from the question to the booked visit
• Clinical guidance stays with the doctor, the assistant guides and books
• Behind it, an analysis of the conversations that says where to invest

Two doors, the same answer

Patients can type in the chat or speak, as they would on the phone. On the other side they find the same assistant with the same information: which test they need, how to prepare, what they can eat beforehand, what it costs, who the specialist is and which days they see patients. The answers come from the centre's own sources: price lists, timetables, preparation instructions.

All the way to the booking

An assistant that only informs leaves the patient halfway. This one goes further: once the person has understood what they need, the assistant checks the visit calendar, offers available slots and sets the appointment. The conversation ends with a date in the diary.

The assistant does not go into clinical guidance, and it says so clearly. It guides and it books. Diagnosis stays with the doctor.

The second layer, the one nobody asks for

Every conversation with a patient is data, and it usually ends up in an archive nobody reads. Here, under the assistant, sits a second layer of AI, a kind of second brain for the centre, that reads the conversations as they come, raw, and turns them into direction. Which visits are asked for most often, where patients get stuck before booking.

How it is built in Microsoft 365

Rebuilt with Microsoft tools, the project would have these pieces. An agent in Copilot Studio, published on the website and on the voice channel, with the centre's sources as its knowledge source. A connector to the visit calendar for bookings. Conversations collected in Dataverse and turned into a Power BI dashboard for management.

In ninety days operational efficiency grew by 60%. Routine questions no longer go through the phone and the front desk, staff have time for the patients who really need a person, and waits have become shorter.

The biggest difference, though, came from the analysis. When the centre had to choose which services to push and what to explain better, it had patients' real requests in front of it.