This thesis explores the transformative potential of Generative Artificial Intelligence (GenAI) in automating financial risk reporting within the banking sector. As financial institutions struggle with increasingly complex risk management requirements, tradi- tional methods of report generation—often reliant on manual effort—face challenges related to efficiency, accuracy, and scalability. This study proposes an innovative AI- driven approach to address these limitations by leveraging advanced machine learning models, particularly those based on Natural Language Processing (NLP) and deep learning architectures. The research outlines the design and implementation of an AI-powered Financial Risk Reporting Automation Tool, which integrates static, semi-static, and dynamic com- ponents to streamline the documentation process. By employing a template-based methodology and placeholder-question mapping, the tool minimizes human interven- tion while maintaining regulatory compliance and contextual accuracy. Through sophis- ticated prompt engineering and automated reasoning pipelines, the system generates high-quality analytical content tailored to specific reporting needs. The results demonstrate significant improvements in operational efficiency, error reduc- tion, and consistency across risk reports. The AI-generated content not only aligns with institutional standards but also provides insightful analysis, enhancing decision- making capabilities. Despite the evident strengths, the study acknowledges challenges such as domain-specific constraints, the need for human oversight, and data privacy considerations. This thesis contributes to the evolving discussion on AI applications in financial services, offering a scalable and adaptive framework for automating risk management documen- tation. The findings pave the way for future research on expanding AI-driven solutions to other areas of financial reporting and regulatory compliance.
Questa tesi esplora le potenzialità trasformative dell’Intelligenza Artificiale Generativa (GenAI) nell’automatizzazione del Financial Risk Reporting nel settore bancario. In un contesto in cui le istituzioni finanziarie devono rispondere a requisiti di risk management sempre più complessi, i metodi tradizionali di produzione dei report, spesso fondati su processi manuali, si trovano ad affrontare sfide crescenti in termini di efficienza, accuratezza e scalabilità. Questo studio propone un approccio innovativo basato sull’IA per fronteggiare tali limitazioni, sfruttando modelli avanzati di machine learning, tra cui quelli basati sul Natural Language Processing (NLP) e sulle tecniche di deep learning. In particolare, vengono presentate la progettazione e l’implementazione di uno stru- mento di automazione del Financial Risk Reporting basato sull’IA, che combina com- ponenti statici, semi-statici e dinamici per ottimizzare il processo di documentazione. Attraverso una metodologia template-based e un placeholder-question mapping, lo stru- mento riduce significativamente l’intervento umano, garantendo al contempo confor- mità normativa e accuratezza contestuale. Grazie a tecniche di prompt engineering e pipeline di ragionamento automatizzate, il sistema genera contenuti analitici di alta qualità, adattati alle specifiche esigenze di reporting. I risultati evidenziano miglioramenti sostanziali in termini di efficienza operativa, riduzione degli errori e coerenza tra i report. I contenuti generati dall’IA non solo rispettano gli standard istituzionali, ma forniscono anche analisi approfondite, migliorando le capac- ità decisionali. Tuttavia, oltre agli evidenti vantaggi, l’approccio presenta anche dei limiti legati alla conoscenza specifica del settore, alla necessità di supervisione umana e alla tutela della privacy dei dati. Con questa tesi, si intende contribuire al dibattito sulle applicazioni dell’IA nei servizi finanziari, offrendo un framework scalabile e adattabile per automatizzare la docu- mentazione nell’ambito del risk management. I risultati ottenuti aprono la strada a future ricerche volte ad estendere le soluzioni basate sull’IA ad altre aree del reporting finanziario e della conformità normativa.
Generative AI: reshaping the path to automation of financial risk reporting in the banking sector
YANG, SOFIA
2023/2024
Abstract
This thesis explores the transformative potential of Generative Artificial Intelligence (GenAI) in automating financial risk reporting within the banking sector. As financial institutions struggle with increasingly complex risk management requirements, tradi- tional methods of report generation—often reliant on manual effort—face challenges related to efficiency, accuracy, and scalability. This study proposes an innovative AI- driven approach to address these limitations by leveraging advanced machine learning models, particularly those based on Natural Language Processing (NLP) and deep learning architectures. The research outlines the design and implementation of an AI-powered Financial Risk Reporting Automation Tool, which integrates static, semi-static, and dynamic com- ponents to streamline the documentation process. By employing a template-based methodology and placeholder-question mapping, the tool minimizes human interven- tion while maintaining regulatory compliance and contextual accuracy. Through sophis- ticated prompt engineering and automated reasoning pipelines, the system generates high-quality analytical content tailored to specific reporting needs. The results demonstrate significant improvements in operational efficiency, error reduc- tion, and consistency across risk reports. The AI-generated content not only aligns with institutional standards but also provides insightful analysis, enhancing decision- making capabilities. Despite the evident strengths, the study acknowledges challenges such as domain-specific constraints, the need for human oversight, and data privacy considerations. This thesis contributes to the evolving discussion on AI applications in financial services, offering a scalable and adaptive framework for automating risk management documen- tation. The findings pave the way for future research on expanding AI-driven solutions to other areas of financial reporting and regulatory compliance.File | Dimensione | Formato | |
---|---|---|---|
2025_4_Yang.pdf
accessibile in internet solo dagli utenti autorizzati
Dimensione
5.65 MB
Formato
Adobe PDF
|
5.65 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/234598