ALGORITHM FOR IMPLEMENTING AI IN THE MANAGEMENT LOOP OF SMES: FROM PROBLEM FORMULATION TO DAILY OPERATION

The paper proposes a practice-oriented algorithm for integrating artificial intelligence into the management decision-making loop of small and medium-sized enterprises (SMEs), covering the full solution lifecycle from managerial problem formulation and pilot selection to deployment, monitoring, and scaling in routine operations. The algorithm is grounded in CRISP-DM and extended with AI Canvas for business design, organizational digital readiness assessment, and an MLOps layer for sustainable model maintenance. The approach is operationalized through seven sequential stages with explicit deliverables, role allocation, and gate criteria, which mitigates typical SME implementation risks (data scarcity, limited skills and budgets, and change resistance). The feasibility is illustrated through an SME inventory-demand forecasting pilot, where a gradient boosting model outperformed a baseline scenario and was embedded with a human-in-theloop mechanism for critical deviations. The results emphasize that combining explainability (XAI) with operational quality controls (data drift monitoring, retraining routines, and usage regulations) improves user trust and the reproducibility of managerial impact.