Research indexSystematic review
АВТОМАТИЗАЦІЯ ПРИЙНЯТТЯ РІШЕНЬ У БАНКІВСЬКІЙ СФЕРІ НА ОСНОВІ АНАЛІЗУ ДАНИХ ВЕЛИКОГО ОБСЯГУ ТА ГЕНЕРАТИВНОГО ШТУЧНОГО ІНТЕЛЕКТУ: ОГЛЯД СУЧАСНИХ ПІДХОДІВ
Гладуш, Клименко, Mazurenko
System technologies30 Apr 2026Open access
NobleIDark:/48914/w1/20261105/C8E42C29DOI10.34185/1562-9945-3-164-2026-12
АВТОМАТИЗАЦІЯ ПРИЙНЯТТЯ РІШЕНЬ У БАНКІВСЬКІЙ СФЕРІ НА ОСНОВІ АНАЛІЗУ ДАНИХ ВЕЛИКОГО ОБСЯГУ ТА ГЕНЕРАТИВНОГО ШТУЧНОГО ІНТЕЛЕКТУ: ОГЛЯД СУЧАСНИХ ПІДХОДІВ
Гладуш, Клименко, Mazurenko · System technologies · 2026
NobleIDark:/48914/w1/20261105/C8E42C29DOI10.34185/1562-9945-3-164-2026-12
Abstract
This paper provides a systematic review of modern concepts, methods, and tools for automating decision-making processes in the banking sector based on generative artificial intelligence technologies. The rapid evolution of the financial sector and the necessity of managing massive datasets have made the integration of intelligent systems a key factor for maintaining competitiveness in modern digital services. The research analyzes the possibilities of using various intelligent technologies, including generative language models, predictive systems, and advanced control automation tools. Special attention is given to the shift from classical machine learning, typically used for risk assessment, to Generative Artificial Intelligence for processing unstructured information in real-time. The study identifies key implementation challenges that remain primary barriers to adoption: model "hallucinations," data privacy concerns, automation bias, and the critical need for strict regulatory compliance. To address these issues, the paper substantiates the expediency of using a multi-level architecture for intelligent systems. This architecture includes the stages of automated data collection, semantic data compression of content, and subsequent classification using vector representations and embeddings. Moreover, the research systematizes specific metrics for evaluating generative artificial intelligence models within the financial sector, emphasizing model resilience, and the level of confidence in decision-making. Finally, the study identifies significant gaps in existing research, particularly the lack of end-to-end machine learning pipelines for the verification of merchant category codes based on website content and outlines the space for the further applied research in the field of merchant activity classification.