Controlling of service maintenance technological process in a 3-tier stochastic transportation system is studied. Managers in the lower tiers of this system have private interests. So, they can manipulate service maintenance indicators to satisfy their interests. To improve service maintenance, top manager uses classification procedure based on machine learning with supervisor's admonitions. The middle-tier manager (taskmaster) controls the workman directly providing services using an adaptive rationing and promotion procedure. Solutions of the game between top manager, the taskmaster and the workman are studied. A controlling mechanism is designed, in which these solutions provide best service maintenance under stochastic conditions. The application of this controlling mechanism is illustrated by the example of service maintenance of passenger cars.