Generative AI and Advanced Analytics for Financial Modeling, Valuation and Strategic Decision-Making.

Résumé fourni par la source

This paper aimed at reviewing the new role of Generative Artificial Intelligence (GAI) and advanced analytics in financial modelling, valuation and strategic decision-making. It also examined the transformative effects of GAI on predictive accuracy, effectiveness and efficiency of decisions, and governance in corporate and investment finance. A systematic literature review was conducted following the PRISMA guidelines, databases employed included Scopus, Web of Science, ScienceDirect, and Google Scholar. The inclusion criteria was set based on the peer-reviewed publications published in 2022-2025 resulting in the retrieval of 35 high-quality papers which were synthesized. The outcome was divided into three categories, namely, (1) current applications, (2) decision-making effects, and (3) human-AI integration and risks. This review shows that there was a clear transition of econometric models, such as ARIMA and GARCH models, to generative systems, such as GANs, transformer-based LLMs, and diffusion models. These constructions have improved predictive accuracy, velocity, as well as, flexibility to support real-time strategic modelling. They are however limited when it comes to the ability to rationalise and the matter of bias and moral responsibility. Human supervision is however still significant particularly in high-stakes financial circumstances that need interpretive and regulatory supervision. Generative AI is improving the precision, analytical and strategic accuracy of financial systems, but also raises new epistemic and governance issues. To increase institutional trust and ethical legitimacy, responsible integration means transparency, fair and control by the human-in-command. To promote innovation and accountability, financial institutions need to establish multi-disciplinary AI governance boards, implement explainability, and promote AI risk literacy.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.