The increasing use of basic artificial intelligence (AI) technologies has recently triggered an enormous boost in AI applications. This also opens up a variety of new application scenarios in the tax sector, as the latest and now third study "Artificial intelligence in the tax sector" by the international consultancy WTS and the German Research Center for Artificial Intelligence (DFKI) shows.
"The influence of the ChatGPT effect can also be felt in the tax sector. The development of new key technologies is opening up completely new opportunities for our industry. At the same time, it is becoming increasingly easy to access and use AI applications without any prior technical knowledge. We are experiencing a turning point," explains Fritz Esterer, CEO of the WTS Group.
The study identified new AI processes that are already established or will prove to be extremely promising in the future for supporting tasks in control functions. It is clear that the use of assistance systems in the tax area can result in significant increases in productivity and quality improvements.
New technologies, new potential
Large language models (LLM), intelligent robot-assisted process automation (iRPA) and explainable AI (XAI) were examined as central technology areas. In addition, more general topics such as machine learning (ML), classification algorithms, anomaly detection and the use of synthetic data were examined in more detail.
"We looked at a variety of technologies, but generative language models such as those from OpenAI or Aleph Alpha undoubtedly have the greatest potential. They are able to understand texts created by humans and generate new ones. This makes language models ideally suited as daily assistants for tax issues, for example, in order to understand tax regulations and guidelines more quickly. They can also be trained for specific tax disciplines," comments Michel Braun, Partner and Chief AI Officer of the WTS Group.
The development of language models has already made considerable progress, meaning that they can already be used in practice today. Nevertheless, the study makes it clear that many tax departments are still reluctant to use AI technologies.
"Frequent obstacles are budget restrictions and data protection issues. In addition, tax functions must be sufficiently digitized and be able to provide high-quality data. Nevertheless, the entry threshold is now very low, and almost all participants in the study agreed that AI will establish itself in the tax sector sooner rather than later," says Prof. Dr. Peter Fettke, scientific director of the study at DFKI.
Wide range of tax applications
The study not only investigated the potential applications of current AI technologies in the tax sector, but also worked out specific application scenarios in various tax disciplines with practice partners and developed prototypes.
"For example, we carried out an accompanying project with a large multinational trading group for the automated generation of transfer pricing documentation. Our aim was for the AI application to extract relevant data from documents and integrate it into the required reporting structure. We have already achieved very successful results with a language model," reports Michel Braun.
Further application scenarios were developed for the VAT classification of service descriptions of invoices, the extraction of information from property tax assessments and the creation of transfer pricing benchmarks.
"Compared to our first two studies from 2016 and 2017, this latest study shows a significant leap in development in the AI environment. We are on the verge of a breakthrough, but this can only succeed if tax functions undergo a transformation. This means that tax managers must also develop skills in dealing with these new technologies. Because it is obvious that AI can only work in cooperation with the human factor," summarizes Fritz Esterer.
The study itself and further information can be found on the AI study page
Representatives of the press have the opportunity to hold background discussions with the authors of the study. If you are interested, please contact florian.kestler@wts.de.
The study was conducted between June 2023 and March 2024 on the basis of technology demonstrators and prototypes. The benefits were determined with the help of expert interviews and workshops with external practice partners and experts from WTS.
About the DFKI
The German Research Center for Artificial Intelligence (DFKI) is Germany's leading business-oriented research institution in the field of innovative software technologies based on artificial intelligence methods. In various research areas and research groups, product functions, prototypes and patentable solutions in the field of information and communication technology are developed on the basis of application-oriented basic research. DFKI is involved in numerous committees for Germany as a science and technology location and enjoys a high reputation far beyond Germany in the training of young scientists.
Further information can be found at www.dfki.de
