Accounting Data and AI for Sustainable Management Decisions: A Conceptual Model and Scenario Analysis

Decisions on agricultural biomass use require the integration of accounting data, technical constraints, market conditions, and the economic value of energy substitution. This study aims to substantiate a conceptual model for supporting sustainable management decisions that combines accounting data, artificial intelligence (AI) forecasting, and linear optimization, and to test its optimization component through scenario analysis. The methodology includes a thematic review of international research, data-provenance mapping, single-period linear programming, and sensitivity analysis using aggregated indicators from an agricultural enterprise for 2024–2025. The case examines the allocation of cereal straw among direct sales, processing into solid biofuel, and internal energy use. The model distinguishes verified accounting and operational facts from management assumptions, derived indicators, and forecast parameters, thereby linking each material input to its source, availability date, and role in the optimization problem. Under baseline conditions, 4,750 t remain for allocation from a maximum available volume of 5,000 t after 5% losses. The calculated solution allocates 4,000 t to internal energy use and 750 t to processing, with an economic objective value of UAH 25.475 million. Scenario analysis shows that the recommended allocation changes in response to losses, the finished-biofuel price, and the economic value of energy substitution. Switching thresholds are UAH 7,368.42/t for the biofuel price relative to direct straw sales and UAH 6,500/t for the value of energy substitution relative to processing, with other parameters unchanged. The contribution lies in formalizing a traceable link between accounting information, forecast parameters, and optimization-based management choices. The objective value is not accounting profit, environmental effects were not measured, and the AI module was not trained or independently tested. Future validation therefore requires dated observations, forecast-accuracy testing and comparison of realized economic decision outcomes.