September 22, 2022

Smart voice assistants conquer many industries – but mumbling doesn’t help

We all know the situation: We call a service number and an automated voice answers - offering to help us if we provide some information first. How do these voice assistants work and what are the challenges? The German AI project SPEAKER is working to simplify the human – machine interaction in a professional environment.
Antonia Kuster

In close cooperation with other technology partners, the Fraunhofer Institute for Integrated Circuits (IIS) and the Fraunhofer Institute for Intelligent Analysis and Information Systems (IAIS) provide customizable technologies. Companies test them in real-world use cases. They are exploring how to make existing technology easier accessible and adaptable for different functions – while securing data sovereignty and data security.

We talked with the representatives of six use cases to find out what is happening.

Is AI now doing taxes?

DATEV is a software house and IT service provider for tax consulting, auditing and the legal profession and offers a wide range of financial services for companies of all sizes. Where software is used, customer questions arise, many of which are simple, but some of which are very specific.

A service hotline is necessary. To relieve staff of many of the simple and repetitive calls, standard cases are answered by voice assistants. This can be up to 20% of all inquiries. This way waiting times can be reduced. For the tricky questions that the AI-based system cannot adequately solve, a real human is always available. Mumbling doesn’t help – we all know that already from experience. But technical terms are fed into the voice assistant, making it possible to train the AI to the specific needs of the company.

Other companies also have call centers, but DATEV’s expertise lies in translating this into German business. Their assistant can answer service and technical questions – support that is urgently needed because DATEV offers a growing number of software products. As a result, the service volume is also rising steadily.

Most important is data sovereignty: Sensitive data may well be included in the requests, and all clients must be able to trust the service 100%. Alexa and other existing voice assistance programs from the US and China have a different understanding of data protection than we do in Europe. One of SPEAKER’s unique selling points was to create a German alternative that upholds European data protection.

Hands-free support for train car maintenance

Very different are the needs of Deutsche Bahn AG: Voice assistants are used for quality assurance and documentation for the maintenance of railcars. The maintenance workers use their hands and get them dirty. At the same time, the repair work requires careful documentation. To support employees, a voice assistant can be of tremendous help – to “take notes” in the process, but also to access the information needed for the repair. The challenges here are the noisy environment and the fact that experienced workers are at times skeptical of digital technology and unfamiliar with it. This always requires a certain amount of persuasion – and of course a minimum level of quality and performance from the prototypes.

Deutsche Bahn has a big workshop factory in Dessau where traction units are completely dismantled in the course of regular maintenance. This is the environment of the SPEAKER use case. A prototype was built to test the workability and find out what is needed to create a smoother process. Worker participation is voluntary at this stage. In the end, two questions need to be answered: Is it worth the cost and effort for the company? And is there added value for the users?

What is the role of data security and data sovereignty here? Everything that is documented must be 100% correct. The train cars transport people and goods, they need to be safe. Mistakes could be dangerous, so trust in the data is very important. Also, the company has competitors who should not get hold of internal data.

Classified car testing

Competition is also an important factor for the tech solution provider IAV. They are very drawn to the SPEAKER project because of its high security standards, a clear priority for the company. As a partner of major OEMs such as Volkswagen and BMW they handle sensitive client data. When testing new cars voice assistants are a big help. The expert tester has only two hands that should remain firmly on the steering wheel and not try to take notes on the side. Basically, IAV says, many jobs that require two hands can benefit from voice translating systems.

The person testing the car can use voice assistants to take notes and control testing software while driving. And convincing the user is always crucial for success. If there are inhibitions, they need to be addressed. Another important part of the process is that the voice assistant is always learning: specialized vocabulary is trained to the AI “on the job”.

Next week we’ll tell you how voice assistants are used in health care, hotels and at live events. Stay tuned!

To learn more about the project, visit:

Author: Antonia Kuster
Antonia Kuster is Digital Communications Manager at IDSA

Stay updated with us