AI Expert Team supports research on the impact of the CSRD on SMEs

The European directive CSRD (Corporate Sustainability Reporting Directive) requires large companies to report on sustainability in addition to financial performance. Although SMEs are not directly subject to this directive, they are affected indirectly. Large companies covered by CSRD must provide insights into the sustainability of their entire value chain and request sustainability data from their suppliers, including SMEs.

What does the CSRD mean for SMEs?

The New Finance research group at THUAS conducted a study commissioned by the Ministry of Economic Affairs on how CSRD impacts SMEs. The study found that 24% of Dutch SMEs are required to provide sustainability data to customers and/or suppliers. For companies with more than 50 employees, this percentage rises to 65%.

As a result, SMEs are increasingly being asked about their CO₂ emissions, energy consumption, and other environmental aspects. However, many entrepreneurs struggle to find the right IT tools to collect sustainability data effectively, leading to an increased workload. At the same time, CSRD presents opportunities; companies that proactively adapt to these reporting requirements can strengthen their competitive position.

AI as an accelerator for sustainability reporting

The AI Expert Team at THUAS played a key role in the research, supporting both quantitative and qualitative analyses. Traditional statistical methods were used for the quantitative analysis, while Generative AI tools were developed to support the qualitative analysis.

How AI was used

AI tools analyzed anonymized interview transcripts and coded statements based on a predefined codebook. A total of 46 interviews were conducted with SMEs, industry organizations, and other stakeholders.

The tools ran locally on work laptops and utilized Generative AI models within The Hague University’s Azure environment, ensuring security and privacy.

AI was not used to replace researchers but to support them. Human assessment and validation remained essential at all times.

Key lesson from this project

It is more effective to have Generative AI models work step by step rather than performing a large number of tasks at once. By breaking tasks into smaller steps, the collaboration between humans and AI remains more transparent and ensures higher quality results.

Want to learn more?

Read the full report here.

Get in touch

This project was carried out by members of the AI Expert Team, including Arie-Willem de Leeuw and Samuel Kernan Freire. Reach out via [email protected].