The article examines how an intelligent decision support system (DSS) becomes part of the learning activity of a management student and under what conditions it develops analytical thinking rather than replacing it. A managerial decision is decomposed into six phases: problem recognition and framing, data collection and analysis, generation of alternatives, evaluation and choice, justification, reflection on the result. Each phase is matched with a component of managerial competency formed within it, and the paper shows how this process changes when part of the operations is taken over by an intelligent system. A model of cognitive load distribution between the learner and the DSS across the decision phases is proposed, with a verification point at which the initiative of analysis must return to the student. The educational effect of working with such systems is described through the opposition of two mechanisms: cognitive amplification, in which the system frees up resources for substantive analysis, and cognitive substitution, in which a ready-made recommendation displaces independent analytical work. It is shown that what switches the system between these modes is not its technical characteristic but the way the learning task is organized and the student’s level of analytical literacy. Instead of a chronological staging, three principles for designing learning situations are proposed: managed incompleteness of the system’s output, mandatory parallel verification, and explicit reflection on the choice procedure. Didactic conditions under which an intelligent DSS works toward competency formation are formulated, and directions for empirical validation of the model are outlined.