MUNICH—Julia Voss works as a dispatcher and firefighter for the Munich Fire Department.

She spent six years before that as a paramedic, riding in the ambulances that she now dispatches throughout the metropolitan area to treat heart attacks, rescue car accident victims and deliver babies, among other things.

A mother of two small children, Voss works one 24-hour shift per week. During that time, she’ll sit at one of the dispatch desks, with its seven computer screens, for three separate three-hour stints. The phone never stops ringing. The pressure is intense; lives are at stake.

“We are constantly on the line,” said Voss. “We provide first aid over the phone until the emergency services arrive, we instruct people in CPR until the emergency doctor is there, we give advice to parents of sick children, we also look after mentally ill people on the phone.”

At the same time, she and the other dispatchers are responsible for arranging transportation for patients who are moving between hospitals or being discharged. The Munich Fire Department sought a way to relieve pressure on the dispatchers while also making the process easier for the nurses and other health-care workers who arrange the transports. The hope was also to shorten waiting times at the hospitals and nursing homes that need the service and to get patients home more quickly. In addition, many of the health-care workers arranging the transports speak German as a second or third language.

The solution? IT experts from the fire department and Microsoft created an AI operator that could handle non‑emergency calls using natural language—in several languages.

They combined several components to build this AI operator.

Microsoft Foundry is where the chatbot’s intelligence is assembled and governed, helping manage how the system responds and what it is allowed to do. Azure Speech (HD Voice), part of Foundry Tools, gives the system a natural‑sounding voice, adjusting tone and cadence based on the words it is speaking. Foundry’s Azure AI Search validates addresses against a municipal database, confirming details such as the correct entrance or other critical information.

Together, the components ideally allow the AI operator to sound natural, stay within its assigned task and verify important details—while leaving human dispatchers in control.

The AI operator can eliminate wait times for those who call in requests while unburdening the dispatchers of the non-emergency calls—giving them more time for medical emergencies and fires and perhaps giving them a little time to take a few deep breaths between calls.

In beta testing, the AI operator has been very effective and easy to use, testers have said, but as Mathias Duensing, one of the firefighters who helped create and advocate for the system, said, “there is always a human in the loop.” He and others at the department emphasized the AI operator is there to improve services; it can’t do what Julia Voss and the other dispatchers do.

“We came up with a saying, ‘Der Bot hilft, der Mensch rettet,’ it means the bot helps but the human saves,” said Duensing, who is head of IT-Architecture.

“If the chatbot doesn’t understand something, or the nurse gets frustrated, they are connected to a human dispatcher immediately.”

In the effort to address the problem, the fire department drew on one of its strengths—the multiple skillsets of its firefighters and dispatchers.

Duensing, for example, has a degree in electrical engineering and information technology and works most days on IT projects, but still serves monthly shifts fighting fires, like Julia Voss, the dispatcher. Florian Dax, one of the other architects of the solution, devoted his university studies—through to a Ph.D.—on subjects related to emergency response services, but he’s not just an academic. He worked as a paramedic and as a dispatcher, too.

He became part of the fire department’s tech team in 2022. “The overlap with IT came mainly through Mathias Duensing, who said: ‘We’re missing someone who can bring IT knowledge into practice—someone who understands what the control center needs, where the pain points are,’” he said.

Duensing and his immediate supervisor, Dr. Tobias Erb, who is also the deputy head of the data center, worked closely together as a team to establish both the staff and technical foundation needed to implement innovative ideas such as the speech bot. Together with Christian Schnepf, the head of the Information and Communication Technology Department at the Munich Fire Department—and, like Duensing and Erb, an active firefighter—the team collaborated with Microsoft to develop the concept of a speech bot, initially aimed at overcoming language barriers.

At many nursing homes and smaller hospitals, the nurses and others handling transport calls come from Eastern European or Asian countries and speaking German on the phone could still be challenging. In 2023, the idea of a natural language AI dispatcher emerged, with the idea of handling non-emergency patient transports through a special line.

“Because I originally came from emergency medical services, I also know how hospitals work: what problems clinics face, what a nurse does all day, how much pressure they’re under,” Dax said. “All of this knowledge flowed into the voice bot.

“For example, when a patient presses the emergency button, the nurse has to run there immediately—but sometimes she’s on the phone with us ordering a patient transport. She then puts us on hold, which slows down everything. These delays cost us a lot of time. So, the voice bot is designed to handle exactly those cases and free up capacity.”

The AI operator will move into broader, real-life testing later this year in the emergency department at LMU Klinikum, Munich’s biggest hospital. Data privacy is governed by European law, and the fire department has been in contact with the data protection authorities throughout the project, Dax said. The fire department will begin working with real patient data at LMU Klinikum in February and soon be able to work with real patient data, he said. Until now it has been using only made-up patient data.

There, as at the fire department, time is a precious commodity. The clinicians at LMU Klinikum have been collaborating on the development of the AI operator.

The call to arrange the transport “might not take that long, but the waiting line often takes a lot of time,” said Dr. Matthias Klein, who leads the hospital’s emergency ward and is also a professor. While nurses are waiting for the dispatcher to take the call, they can’t do much else, he said, because they must be ready when someone eventually picks up. If there is an urgent situation on the emergency ward, the nurse will have to hang up and try again later.

“It’s really a time-consuming process at the moment,” he said. “And it’s really an important moment right now with technology and medicine. Medicine is often at the forefront of technology, but AI is really coming in a lot of ways right now.”

He said the AI operator at the Munich fire department has the potential to help open hospital beds more quickly.

“We have the patient ready [to leave the hospital], we did everything that’s necessary medically, but then we have this obstacle—the transport,” he said. “That starts with ordering the transport, which already takes some time, and then we often have to wait for the pickup. And that can take several hours. And if it stretches into the evening, we might have to cancel because we don’t want the patient to arrive at the final destination at 1 a.m.”

Klein said that from the hospital’s perspective, it will save nurses’ time that can better be devoted to patient care. He hopes it will eventually help improve the efficiency of the transport system, to open hospital beds more quickly by cutting wait times.

Florian Dax said that it’s hard to understand the emotional stress that the dispatchers face each day.

One minute they might be trying to explain to a panicked parent how to perform CPR on a critically ill child. The next dealing with someone’s heart attack or seizure in real time. Meanwhile the lights are flashing on their screens, and another call is coming in.

For Dax and the Munich Fire Department, the most important thing is using any tool to improve its life-saving role.