Risks of applying generative artificial intelligence in BI technologies

Abstract

This article examines the risks of applying generative artificial intelligence in modern BI technologies. It is shown that integrating generative AI into Business Intelligence systems opens up new opportunities for automating analytical reporting, generating natural language queries, generating program code, forecasting, and supporting management decisions. However, the use of such technologies is accompanied by significant threats related to the unreliability of analytical conclusions, data quality, information security, confidentiality, opaque algorithms, legal and ethical restrictions, organizational dependence on suppliers, as well as the risks of cyberattacks and analytics manipulation. Particular attention is paid to the problem of drawing convincing but factually incorrect conclusions that can lead to erroneous management decisions. Based on an analysis of scientific literature, approaches to AI risk management, and BI system implementation practices, a classification of generative AI risks in BI is proposed and mitigation strategies are identified. These strategies include data quality management, access control, request and response auditing, the use of verifiable results, data localization, model testing, the implementation of the human-in-the-loop principle, the development of regulations, and user training. A conclusion is reached regarding the need to view generative AI in BI not as an independent source of truth, but as an auxiliary tool requiring control, transparency, and responsible use.