(Partial) automation of online customer service

Development of an intelligent digital agent for (partially) automated replies to customer enquiries

Project team

Prof. Dr. Mathias Klier
Patrick Bedué
Eleonore Graef
Roland Graef
Johannes Redl
Jan-Felix Zolitschka



Due to a strong increase in the number of users in online forums, it is a great challenge to answer all customer enquiries in a timely, personalised manner and with consistently high quality. Companies must therefore look for new ideas and solutions that utilise new methods and technologies to support the answering of customer enquiries in the forum.

Ulm University, in cooperation with a large German insurance company, took on the project of developing an intelligent digital agent that supports insurers in making their online service more efficient and customer-oriented. The partners have developed a new, innovative tool that supports online service employees in answering customer enquiries. The aim of the project was to investigate to what extent it is possible, based on the latest research, to revolutionise customer service for the benefit of the company and its customers in and thus achieve the next stage of digital transformation.

The intelligent digital agent is designed according to the approach of case-based reasoning. This is an artificial intelligence methodology that successfully solves new cases based on similar previous cases. This methodology enables the digital agent to use the knowledge from questions previously answered in the forum as well as from an additional insurance-specific knowledge base to automatically suggest replies to new customer queries. In order to ensure consistent quality despite automated reply suggestions, the employees check suggested answer options for correctness before they are made available to the customer via the forum. Correctly solved customer enquiries are then transferred to the knowledge base of the digital agent. This grows the knowledge base with each additional question solved in the forum and the digital agent improves in a self-learning manner. By combining case-based reasoning with a methodology specifically developed for online forums, an approach was designed that is able to learn as well as flexible and also significantly improves efficiency in the workflow.

The results of the project show: The quality of the newly developed methodology already exceeds all previous approaches by at least 25 percentage points with the first two reply suggestions. This allows employees to devote more of their working time to particularly demanding customer enquiries and to answer them efficiently with the support of artificial intelligence. In addition, the digital agent ensures a faster and more consistent response to customer enquiries, which further improves the customers' perception of service response time and service quality.

Cooperation partner: German insurance company

Project period: July 2017 - January 2018


This pilot project demonstrates how close collaboration of researchers and practical experts and managers can result in the rapid development of a revolutionary prototype for practical use while laying the scientific foundation for promising research activities. While the researchers at the Institute of Business Analytics were able to gain visibility and reputation through corresponding research publications, the text-based digital agent developed within this framework serves as the basis for the company-wide implementation of a high-quality solution.