ESG criteria are becoming increasingly important in the financial services industry (FSI). Banks and financial institutions are faced with the challenge of efficiently collecting and processing ESG data and using it to fulfil regulatory requirements (such as the EU Taxonomy and the Corporate Sustainability Reporting Directive, CSRD). To address these requirements, the flagship project „AI-based ESG assessment for the FSI“ was launched as part of the expert committee AI for FSI initiated and in co-operation with our partners Yukka Lab and Deloitte implemented. In a two-stage series of workshops, key ESG challenges for banks were identified and suitable AI-based solutions were developed to demonstrate concrete implementation strategies. 

Workshop series: Overview 

The flagship project was divided into two workshops. In the first workshop, we focussed on the Use Case Ideation where key ESG challenges were identified together with representatives of the banks. The second workshop served to find solutions, in which two promising areas of application were analysed and discussed in detail. 

 Results of the first workshop: Use Case Ideation 

After intensive and individualised work with the various banks, we developed five use cases in the first workshop: 

1. Automated ESG data integration The aim is to collect and integrate ESG data from various sources in order to enable more comprehensive analyses. Data quality and availability have been recognised as critical challenges. 
   
2. Efficient ESG reporting The automation of ESG reports should reduce the workload and improve consistency. It became clear that large amounts of data need to be processed and contextualised efficiently. 
   
3. Document processing for energy certificates The focus is on the automatic reading and processing of energy performance certificates in order to reduce manual work and speed up the process. The different formats of the documents were cited as a particular challenge. 
   
4. ESG scoring platform A tool for evaluating a wide range of ESG data was identified as a potential use case, with the availability of information and validation of the data being cited as a challenge.
   
5. CSRD report creation The aim is to simplify complex reporting obligations and reduce the resources required. Challenges include the large number of requirements, necessary infrastructure changes and changing guidelines. 

The findings of the first workshop show that the availability and quality of ESG data are crucial for the successful implementation of AI-based solutions. It also became clear that process automation can increase efficiency and reduce manual errors, while compliance with regulatory requirements remains central. 

Results of the second workshop: Finding solutions 

In the second workshop, concrete solutions were to be developed. In order to go into greater depth, we worked with our partners Yukka Lab and Deloitte to identify the two application areas from the first workshop that were most relevant to the participants. These two application areas were then analysed in more detail together: 

1. ESG data platforms
   The discussion centred on the integration of ESG data from various sources and the technical feasibility of such platforms. One finding was that automated and semi-automated data imports are possible through tools such as Yukka Lab’s. Currently, the available data is mainly based on „inside-out“ data relating to past reporting periods. This limitation should be overcome by a more comprehensive and forward-looking data basis in order to be able to react more proactively to changes. The participants discussed various areas of application, such as analysing minimum safeguards or obtaining data for CSRD reports. The availability and quality of the data remained key challenges that need to be considered separately depending on the area of application. 
   
2. Spatial finance: assessing risks and analysing assets 
   With regard to relevant ESG data, NVIDIA also shows how AI and machine learning can be used in the Spatial Finance can be used to monitor assets, assess risks and analyse insurance claims. By combining geodata with AI-powered tools, banks can better understand the impact of environmental changes on their investments and make informed decisions. These innovative approaches make it possible not only to utilise ESG data efficiently, but also to proactively manage future risks. 
   
3. Document AI
   The focus was on the automated extraction and processing of documents such as energy performance certificates. The participants discussed challenges such as the poor legibility of old documents and the effort involved in manual extraction, categorisation and corrections. It was shown that a high level of automation is possible, replacing up to 90% of manual work. At the same time, it was emphasised that complete traceability and reliable logging are essential to ensure the quality and compliance of the results. 

NVIDIA added to this discussion with its NIM Agent Blueprintswhich were developed for automated document processing. This also allows complex RAG (Retrieval Augmented Generation) pipelines to be set up, which enable structured extraction and processing of information from large volumes of PDF data. Such solutions open up new opportunities to further automate and improve ESG reporting.  

Practical implementation and next steps 

The implementation of the use cases follows a structured approach. In the area of ESG data platforms the process starts with a PoC to evaluate data quality and model adjustments. This is followed by user training and integration into existing IT systems, including the definition of interfaces. 

In the area Document AI the project also begins with a PoC to test the model’s performance. This is followed by integration into systems and processes. Continuous monitoring of the model performance is essential to ensure reliable automation. 

NVIDIA offers companies with LaunchPada comprehensive End-to-end AI platformThe platform enables AI solutions to be quickly tested, developed and productively implemented. For ESG data platforms in particular, the platform allows applications to be quickly commissioned and scaled without the need for extensive upfront investment. This allows banks to realise their projects more efficiently and with less risk. 

Sustainability and energy efficiency through accelerated computing 

Another key aspect when using AI solutions in an ESG context is their sustainability. NVIDIA emphasises the importance of Accelerated Computing as a basis for Sustainable Computing. Through the targeted use of GPU-accelerated technologies, not only can processing speeds be significantly increased, but energy consumption can also be reduced. According to NVIDIA, accelerated computing enables an increase in efficiency of up to 20 times compared to conventional CPU-based systems. This makes the NVIDIA AI Platform a sustainable yet powerful solution for banks that want to meet ESG requirements while minimising their carbon footprint. 

Conclusion 

The workshops have shown that close collaboration within the AI Park ecosystem can identify concrete solutions for ESG challenges. During the workshops, specific use cases were identified at the banks and suitable solutions were developed. These collaborations enable data platforms and automated document processing to be used effectively and existing processes to be optimised. The AI Park provides a platform that promotes exchange and collaboration between leading companies and institutions, thereby supporting the development and application of innovative AI solutions.   

For questions and further bilateral discussions, please contact Yukka Lab, Deloitte and NVIDIA: 

Yukka Lab: 

Sascha Wagner:swa@yukkalab.com  

Deloitte: 

Tobias Waetzig:twaetzig@deloitte.de  

Dr Oliver Fink:ofink@deloitte.de  

NVIDIA: 

Dr Jochen Papenbrock:jpapenbrock@nvidia.com 

Thank you once again for your cooperation and commitment! 

  

Kind regards, 

The Flagship team of the KI Park, Yukka Lab, Deloitte & NVIDIA