Vertical concern is a group of companies with the mills from independent industries under common management. According to the INDUSTRIE 4.0 concept, management of such concern can apply artificial intelligence, including machine learning. Also it is necessary to take into account the human factor. The article explores the problem of using organizational control theory and machine learning to increase production output from the mill operating in one of concern industry. Changes lead to stochastic fluctuations in production output of such mill and industry. The lower-level element of vertical concern hierarchy knows the potential of own production output better than the higher-level element. Therefore, such lower-level element can manipulate its output to get more inducements from the higher-level element as a result of machine learning. Such unwanted activity can result in decline in total production output of the concern. To avoid this activity, the mechanism for production output management is derived including two-level machine learning and inducements. Sufficient conditions are found for the synthesis of such a mechanism in which the stochastic possibilities of increasing of a production output in a concern are used. The implementation of such a mechanism is illustrated by the example of a locomotive refit management in concern Russian Railways.