BUSINESS INTELLIGENCE IN THE CONTEXT OF THE INTELLECTUALIZATION OF ANALYTICAL INFORMATION SYSTEMS

Abstract

Business Intelligence (BI) systems have long served as a primary mechanism for decision support by providing standardized metrics, consistent semantic logic, and stable analytical artifacts such as reports and dashboards. This paradigm enables reproducibility of analytical outputs and comparability of indicators within recurring managerial cycles. At the same time, the intellectualization of analytical systems, the rapid growth and heterogeneity of data sources, and the increasing prevalence of iterative and context-dependent decision questions expose limitations of an artifact-centered BI approach. These limitations emerge across the “data–interpretation” chain, including issues of integration and data quality, the rigidity of predefined analytical logic, the proliferation of analytical artifacts and the rising effort required to maintain them, as well as interpretation gaps between user context and the analytical model. The paper provides a theoretical and analytical examination of BI in the context of intelligent analytical ecosystems and systematizes the key constraints of traditional BI practices with respect to requirements for timeliness, flexibility, and controllability of analytical results. Based on this analysis, the study proposes an author’s hybrid BI–RAG concept in which BI artifacts remain the canonical representation of standardized metrics and act as a “source of truth” for recurring decision scenarios, while retrieval-augmented generation (RAG) is introduced as a conversational interface for accessing analytical information. In the proposed approach, the RAG layer supports contextual explanation, navigation to relevant BI artifacts, and iterative request formalization, while emphasizing an evidence-based interpretation principle grounded in relevant knowledge sources. The scientific novelty of the work lies in formalizing the hybrid interaction model through an actor model, a reference architecture of the BI–RAG setup, and a generalized workflow model for decision-request handling. The workflow representation indicates that introducing a “RAG-first” stage can reduce the number of process steps between a user’s request and access to relevant analytical information in cases where the request can be addressed through explanation or routing to an existing canonical BI view. The results have theoretical significance for the development of approaches to organizing modern analytics ecosystems and practical value for designing hybrid analytics access solutions under data governance requirements and controlled change management.

Keywords: analytical information systems, decision support systems, data analytics, artificial intelligence, retrieval-augmented generation (RAG), data-driven decision-making, analytical ecosystems.