From Documents to Disclosures: Automating Sustainability Reporting with Large Language Models
(2025) INTM01 20251Innovation Engineering
- Abstract
- This study explores the use of Large Language Models (LLMs) to automate sustainability reporting in accordance with the European Sustainability Reporting Standards (ESRS). Given the increasing complexity and regulatory demands of sustainability reporting, organizations face significant challenges connected to their reporting processes. To evaluate the potential of LLMs in addressing these challenges, a prototype system was developed in collaboration with Position Green. The goal of the prototype was to automate the extraction of relevant information from internal company documents and generate initial draft ESRS disclosures. The development process followed an action research methodology, with iterative refinements of the prototype.
The... (More) - This study explores the use of Large Language Models (LLMs) to automate sustainability reporting in accordance with the European Sustainability Reporting Standards (ESRS). Given the increasing complexity and regulatory demands of sustainability reporting, organizations face significant challenges connected to their reporting processes. To evaluate the potential of LLMs in addressing these challenges, a prototype system was developed in collaboration with Position Green. The goal of the prototype was to automate the extraction of relevant information from internal company documents and generate initial draft ESRS disclosures. The development process followed an action research methodology, with iterative refinements of the prototype.
The relevance of the solution was assessed through internal interviews and a survey targeted at sustainability reporting professionals. Empirical findings indicate that while the specific challenges faced by companies vary, most struggle with data collection, time constraints, and interpreting ESRS requirements. The prototype demonstrated potential in addressing these challenges by reducing the manual effort required and enabling more efficient drafting of reports.
However, the prototype also had limitations, including hallucinations, biased answers, limited source traceability, and difficulties in quality assurance. These challenges emphasize the need for human validation together with the system. Despite these limitations, this study demonstrates the value of integrating LLMs as supportive tools in sustainability reporting. With stakeholders’ and consumers’ demand for transparency, LLM-based systems can support both mandatory and voluntary reporting efforts. As AI technologies continue to evolve, their reliability and analytical capabilities are likely to improve. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9204394
- author
- Nystedt, Amanda LU and Wiksten, Oliver LU
- supervisor
- organization
- course
- INTM01 20251
- year
- 2025
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- Sustainability Reporting, Artificial Intelligence, AI, Large Language models, LLMs, CSRD, Corporate Sustainability Reporting Framework, European Sustainability Reporting Standards, ESRS, RAG, Retrieval-Augmented Generation
- language
- English
- id
- 9204394
- date added to LUP
- 2025-06-23 13:51:13
- date last changed
- 2025-06-23 13:51:13
@misc{9204394,
abstract = {{This study explores the use of Large Language Models (LLMs) to automate sustainability reporting in accordance with the European Sustainability Reporting Standards (ESRS). Given the increasing complexity and regulatory demands of sustainability reporting, organizations face significant challenges connected to their reporting processes. To evaluate the potential of LLMs in addressing these challenges, a prototype system was developed in collaboration with Position Green. The goal of the prototype was to automate the extraction of relevant information from internal company documents and generate initial draft ESRS disclosures. The development process followed an action research methodology, with iterative refinements of the prototype.
The relevance of the solution was assessed through internal interviews and a survey targeted at sustainability reporting professionals. Empirical findings indicate that while the specific challenges faced by companies vary, most struggle with data collection, time constraints, and interpreting ESRS requirements. The prototype demonstrated potential in addressing these challenges by reducing the manual effort required and enabling more efficient drafting of reports.
However, the prototype also had limitations, including hallucinations, biased answers, limited source traceability, and difficulties in quality assurance. These challenges emphasize the need for human validation together with the system. Despite these limitations, this study demonstrates the value of integrating LLMs as supportive tools in sustainability reporting. With stakeholders’ and consumers’ demand for transparency, LLM-based systems can support both mandatory and voluntary reporting efforts. As AI technologies continue to evolve, their reliability and analytical capabilities are likely to improve.}},
author = {{Nystedt, Amanda and Wiksten, Oliver}},
language = {{eng}},
note = {{Student Paper}},
title = {{From Documents to Disclosures: Automating Sustainability Reporting with Large Language Models}},
year = {{2025}},
}