The role of Artificial Intelligence in the corporate sustainability reporting in the context of Corporate Sustainability Reporting Directive (CSRD)

Abstract

This thesis examines the challenges and opportunities for Artificial Intelligence (AI) in sustainability reporting in accordance with the Corporate Sustainability Reporting Directive (CSRD). The first step in the literature review was to outline the key applications of AI related to sustainability reporting and then assess the regulatory and ethical aspects of its implementation in organisations. Next, global initiatives such as the UN 2030 Agenda and the Paris Agreement were examined to provide context for the European Union’s policy framework in the field of climate governance. The background material also covers EU sustainability policies such as Taxonomy, the Corporate Sustainability Disclosure Directive (CSDDD), but the focus is specifically on the CSRD.

From a methodological perspective, this exploratory study is based on qualitative data collected from six semi-structured interviews conducted with three AI and data specialists, as well as three practitioners in the field of sustainability reporting. The data obtained from the interviews was systematically coded using MAXQDA, followed by an abductive thematic analysis. This analytical procedure revealed six main themes and eighteen sub-themes. The results highlight the advantages that AI can provide in meeting CSRD requirements, including the automation and structuring of large volumes of unstructured data, as well as reducing the technical barriers faced by reporting teams and specialists. Furthermore, AI has the potential to accelerate the understanding of regulatory provisions by synthesising them and answering complex questions, thereby reducing the learning curve for practitioners.

However, the study highlights several important challenges. These include the risks of bias and data manipulation associated with LLMs. Alongside this is the tendency to over-rely on automated results without sufficient human oversight. There is also the environmental impact of AI systems to consider, given their high energy and resource consumption in data centres. Therefore, during the interview, conflicting opinions on the role of AI in decarbonisation emerged. While AI can optimise energy consumption, its overall environmental impact depends on the energy mix and progress towards renewable energy-powered computing systems.

The study is limited to EU-focused regulations and qualitative data collected during the first ten months of 2025, which restricts its generalisability. It concludes that AI can materially support CSRD compliance when complemented by robust governance, transparency, and liability frameworks, and when technical deployment aligns with decarbonised energy sources. Future research should employ mixed methods and a wider geographic scope to enhance validation and policy relevance.