Application of Generative AI in the context of sustainability reporting: Twin Transformation using the example of Rödl & Partner

published on 24 March 2025 | reading time approx. 4 minutes

Sustainability reporting in Europe has expanded due to initiatives such as the Corporate Sustainability Reporting Directive (CSRD), the EU Taxonomy and the Corporate Sustainability Due Diligence Directive (CSDDD), which poses new challenges for companies. Rödl & Partner uses modern technologies such as Generative AI to meet high-quality standards in its services. Rödl & Partner advances the Twin Transformation, which combines technological development and sustainable business practice, and uses AI co-pilots in which AI and humans work together. Expert bots use Retrieval-Augmented Generation (RAG) in this context to access validated databases in order to deliver reliable results and improve knowledge management.

Sustainability meets Generative AI

In recent years, sustainability reporting has been significantly expanded at European level. Initiatives such as the CSRD, including the associated European Sustainability Reporting Standards (ESRS), the EU Taxonomy Act and the CSDDD pose new challenges for companies in all sectors. This also applies to Rödl & Partner as a provider of high-quality consulting and auditing services in the area of sustainability reporting.

To meet the high-quality standards of offered consulting and auditing services in the field of sustainability reporting, Rödl & Partner uses the latest technologies and, in particular, the capabilities of Generative AI. In doing so, Rödl & Partner actively drives the realization of the Twin Transformation, which defines the union of technological development and sustainable business practice (cf. Gu/Dai/Vasarhelyi, 2023; Jiang/Gu/Dai, 2023).

AI Co-pilot for the realization of the Twin Transformation

Specifically, Rödl & Partner relies on the concept of the AI co-pilot (see initially: Gu et al. 2024). The AI cooperates with the human counterpart to support a wide range of complex activities. Both the AI and the human experts play to their strengths to optimize efficiency and effectiveness. The human experts bring methodological expertise, specialized knowledge, and experience in the specific subject area to solve the problem. The AI model, on the other hand, analyzes large amounts of data, identifies patterns, and provides an initial indication of an interpretation. Additionally, the human experts improve the AI model through regular feedback and simultaneously benefit from the outputs generated by the AI model (cf. Reutter/Föhr, 2025).

Such a concept of an AI co-pilot can be used for various applications. For example, Frank Reutter and Dr. Tassilo Föhr from Rödl & Partner recently comprehensively explained the AI co-pilot in the journal "Die Wirtschaftsprüfung" (WPg) using the example of an IT audit according to ISA 315 (Revised 2019).

This concept of the AI co-pilot can be transferred to a variety of other application areas, particularly to applications in the field of sustainability reporting for the realization of the Twin Transformation.

Generative AI as a basis for knowledge management

Rödl & Partner relies particularly on specialized expert bots that can access databases validated by Rödl & Partner using Retrieval-Augmented Generation (RAG). This increases the knowledge base of the large language models of these expert bots compared to large language models that generate output without a database extended by RAG. RAG is a technology that combines Generative AI models with information retrieval capabilities and can be described using two essential processing steps (cf. Lewis et al., 2020):

Retrieval: In this phase, relevant information is extracted from a large database or document pool based on the initial prompt. This is usually done by using intelligent semantic search algorithms that identify the most suitable documents or text sections based on the question asked or the given context.
Generation: In the next step, this pre-extracted information is used to formulate a coherent and informative answer. A large language model processes these previously retrieved data and creates an answer that considers both the original question and the found information.

Rödl & Partner has various expert bots that can access different and validated expert knowledge in various data formats. The answers of the expert bots always refer to the stored sources, including the reference to the original document from which the expert bot obtained the information. Expert bots that access specialized expert knowledge therefore deliver reliable results. This output can therefore be used to address complex issues. Additionally, human experts can always rely on structured knowledge management in the field of sustainability reporting.

Conclusion

The concept of the AI co-pilot in the area of structured knowledge management combined with comprehensive technical expertise can – as at Rödl & Partner – take the service portfolio of a professional services firm to the next level and actively drive the Twin Transformation forward. However, the Twin Transformation also offers great and promising potential for many other companies in a wide range of industries. Against this backdrop, it is advisable for every company to analyze the far-reaching opportunities, ideally together with proven experts, in a timely manner and to leverage the existing potential.

Further References

Gu, H., M. Schreyer, K. Moffitt, and M. Vasarhelyi. 2024. Artificial intelligence co-piloted auditing. International Journal of Accounting Information Systems 54: 100698. https://doi.org/10.1016/j.accinf.2024.100698
Gu, Y., J. Dai, and M. A. Vasarhelyi. 2023. Audit 4.0-based ESG assurance: An example of using satellite images on GHG emissions. International Journal of Accounting Information Systems 50: 100625. https://doi.org/10.1016/j.accinf.2023.100625
Jiang, L., Y. Gu, and J. Dai. 2023. Environmental, social, and governance taxonomy simplification: A hybrid text mining approach. Journal of Emerging Technologies in Accounting 20 (1), pp. 305–325. https://doi.org/10.2308/JETA-2022-041
Lewis, P. E. Perez, A. Piktus, F. Petroni, V. Karpukhin, N. Goyal, … and D. Kiela. 2020. Retrieval-augmented generation for knowledge-intensive NLP tasks. Advances in Neural Information Processing Systems, 33, pp. 9459-9474.
Reutter, F./Föhr, T. L. 2025. Generative KI als Copilot in der Abschlussprüfung. Die Wirtschaftsprüfung (WPg). 78 (2), pp. 59-67.