We analyze the performance of the agents of organizational-learning oriented classifier systems (OCS) with and without the implementation of XCS decision criterion inside a Pac-Man world. The agents act as the enemies of Pac-Man. Pac-Man is controlled by random movements and several different Pac-Man's artificial intelligence algorithms proposed in other researches. If they capture Pac-Man, the organizational performance is indicated by the number of remaining dots and the amount of time required for capturing Pac-Man. If they fail to capture Pac-Man, the organizational performance is indicated by the amount of time required by Pac-Man to eat all dots.