Using chatbot-based process modelling to accelerate the digitalization of SMEs
Many companies face the challenge of supporting their activities with digital technologies in a targeted manner. Business process models play a major role here —especially when introducing corresponding information systems. However, small and medium-sized enterprises often find it difficult to record and improve their processes. As part of the KICoPro project, a chatbot-based modeling copilot will be developed to help capture and improve business processes by harnessing the power of advanced large language models.
Business process management (BPM) is considered an important basis for achieving corporate goals. Business process models play a central role in the BPM life cycle for designing, controlling and analyzing business activities. Among other things, they serve as the basis for the introduction of information systems, requirements engineering and, especially, companies' digitalization projects.
While large companies have already been able to build up their own expertise in this area, many small and medium-sized enterprises (SMEs) often find it difficult to deal with the topic of BPM in a sufficiently professional manner. Practical experience has shown that many digitalization and optimization projects fail or are not even tackled due to a lack of process-related consideration. Examples of affected sectors include municipal utilities (energy and water supply, wastewater and waste disposal) and the manufacturing industry.
As a result, information systems are often only introduced in a rudimentary way and without sufficient process integration. That means that significant digitalization potential is not exploited. As the Bertelsmann Foundation study from 2020 points out, SMEs are lagging large companies in this respect. Without in-depth modeling of existing processes, there is also a risk of enriching existing — and often suboptimal— processes with digital solutions that fail to address underlying problems (see quote from Thorsten Dirks at the Süddeutsche Zeitung 2015 business summit).
For many SMEs, creating an initial process model to document existing company procedures, which is essential for BPM, is often not feasible due to either a lack of expertise or financial constraints.
As part of the KICoPro project, the potential of large language models (LLM) will be researched and used to support companies, particularly SMEs, document and improve their processes. The aim is to enable companies to record their processes with the help of a chatbot. The LLM-supported chatbot will act as a process modeler, generating accurate process models by asking the right queries. AI should also assist in transforming documented current processes into optimized target processes. This could reduce the barriers for companies to successfully implement digitalization projects. The use of AI could significantly reduce the costs of process data collection as well, so that a broad application in the economy can be achieved. The scientific focus is on the research question of the extent to which existing language models, which have so far been trained exclusively on textual data, have already learned “process knowledge” and can thus take on the role of the process modeler. In addition, new training methods are to be developed and tested to teach language models a better understanding of processes.
The project is funded by the BMBF under the measure “KMU-innovativ” with the funding code 01IS24053C. The consortium consists of Prime Force Freiburg GmbH (development partner), Pentadoc AG (application partner) and the Institute for Information Systems (IWi) at DFKI (research partner). Various smaller companies are involved as associated partners.
A publicly accessible prototype will be made available during the project and announced via the project partners' social media channels.
About Prime Force
Prime Force Freiburg GmbH (PFF) is an IT consulting company with a focus on document-centric processes, which currently employs 5 people at its Freiburg location. The Prime Force Group employs around 150 people in various branches across all locations. The branches are each operated as independent limited companies. The founders of PFF have been working on the topic of AI since 1998 and 2001 respectively, with the success of AI through deep learning and LLMs far exceeding the technologies used up to that point. Digitization and intelligent classification of documents as well as data extraction are PFF's areas of expertise, which have already been proven in large corporations as well as smaller public clients. Manual process design and its manual implementation are some of PFF's specialties.
About Pentadoc
Pentadoc AG is a product-independent management consultancy that has specialized in the digitization of documents and processes since the company was founded in 1998. The medium-sized consultancy has been advising companies in various industries for 25 years. Thanks to its many years of experience, Pentadoc not only has methodological expertise in the field of process consulting, but also in-depth technical and procedural expertise in the relevant core industries and has practical insights into the specific challenges and often existing deficits on the part of SMEs in the area of GPM.
About the IWi at DFKI
The Institute for Information Systems (Institut für Wirtschafsinformatik, IWi) at the German Research Center for Artificial Intelligence (Deutschen Forschungszentrum für Künstliche Intelligenz, DFKI) in Saarbrücken has been working on business process management and process modeling for many years. In particular, the use of artificial intelligence (AI) and the added value it generates for companies is one of the main areas of research. This project draws on experience in the field of process modeling and the innovative use of AI to control and improve business processes. The AI hardware available at the institute ensures the secure processing of company data.