ROLF makes AI fit for business decisions

Using artificial intelligence (AI) to support complex decisions in a business environment - this is the aim of the ROLF research project. The acronym stands for Reinforcement Learning for Business Processes. The project, which is funded by the Bavarian State Ministry for Economic Affairs, Regional Development and Energy and is scheduled to run for two and a half years, is supported by DATEV eG and the Fraunhofer Institute for Integrated Circuits IIS.

Business management and entrepreneurial decisions are usually characterised by a multitude of factors, complexities and dependencies. A decision at one point - be it by the entrepreneur, auditor or tax advisor - often has a major impact on the options available afterwards and also on the long-term results. Here it would be beneficial if decision-makers were given more relief and support in the future, for example in the form of assistance systems - also against the background of the current labour shortage.

Artificial intelligence has great potential to serve as the basis for such assistance systems. In this context, reinforcement learning is a promising approach to teach such systems strategies (called policy) on the basis of which they can make the optimal, sequential decisions with which a good long-term result can be achieved and an overall goal can be reached. This specific method of machine learning learns autonomously through interaction with an environment via a reward principle that rewards "long-term correct decisions" accordingly.

More about Reinforcement Learning

Better decisions with ROLF

Despite this potential, reinforcement learning has so far led a niche existence in the business environment. This is not least due to the difficulty that business processes are often only loosely documented and the data available for effective training is either non-existent or only available in small quantities. In the ROLF project, DATEV and Fraunhofer IIS are now systematically testing the applicability of reinforcement learning for business management processes in a very practical way in exemplary application scenarios.

Fraunhofer IIS is contributing its experience and comprehensive background knowledge in the field of "experience-based learning". In the project, the researchers are responsible for the basic technology - from the evaluation of the reinforcement learning method suitable for the respective purpose, to the design of the algorithms and the interfaces to the respective required data sources. DATEV takes over the coordination and project management, defines the technical requirements and ensures the adaptation to the specific technical context. The IT service provider involves both internal AI experts and specialists from various business fields, including auditing and law firm management. In addition, the law firm Christoph Renz, with a view from the perspective of tax consultancy, and the company Celonis, as a specialist in the digital mapping of (business) processes, will accompany the research project in an associated manner.

If you have any questions about the project or would like to exchange information on the topic, please contact Dr Eva Hörster at DATEV at forschungskooperationendatev..de.