Of all the services that AI can provide in the fight against global warming and decarbonization, a concrete and operational use case for businesses stands out: AI can make it easier for businesses to comply with the CSRD directive. As we know, the European Commission’s aim with this set of rules is to harmonize companies’ sustainability reporting and improve the comparability and quality of published ESG data. This is a very positive step in principle, but it could prove to be a headache for companies. Compared with the previous directive of 2014 (Non-Financial Reporting Directive - NFRD), the CSRD will gradually apply between 2024 and 2028 to a much larger number of companies (including all companies listed on European regulated markets). It imposes standardization for reporting obligations, the publication of a sustainable development report in a compulsory digital format as part of the management report, and the compulsory verification of information by a statutory auditor or an independent third-party body.
As a result, many French companies are now required to publish detailed information on their environmental, social, and governance risks, opportunities, and impacts. The cornerstone of the CSRD is ‘double materiality’, in other words, the obligation for companies to publish information on the impact of sustainability issues on their financial performance, but also on their own impact on the environment and society. In total, almost 1,200 indicators must be provided, in the form of figures or written content, on a wide range of subjects such as the level and scale of its greenhouse gas emissions and the efforts made to reduce them, as well as working conditions, diversity, inclusion and the role and operation of its governance bodies.
AI and generative AI tools will therefore be essential throughout the process of gathering, analyzing, and formatting information. An AI tool can, for example, download a huge amount of a sector’s extra-financial reports to help identify benchmarks and key issues, without the need for exhausting and time-consuming research. Generative AI can also pre-write the content, since some CSRD indicators are to be written, and generate the first versions of the sustainability report based on the corpus of data from ‘healthy’ companies on which it has been trained.
Thanks to ‘fuzzy matching’, an algorithmic process based on the approximate correspondence of data, it will be possible to process the company’s huge amount of data and ‘pre-wire’ it to the 1,200 indicators of the CSRD. However, this implies putting in place tools to identify reliable, precise, and accurate data, because, as we have clearly understood, the exercise does not simply consist of publishing raw data, but verifiable data. In other words, this means adopting a responsible approach to AI in terms of controlling sources (indicating pdf sources or use cases used, etc.) and making quantified indicators or content explicit, particularly in terms of the company’s preparedness and adaptation to climate change.
CSRD therefore requires a global process for putting the company’s sustainability data and scenarios in order, which only ‘scaled’ and responsible AI can make more reliable, faster and more operational.
Content from Axionable. Article written by Gwendal Bihan, Vice-Chairman of Impact AI and Chairman of Axionable.
