Roles with artificial intelligence (AI) behavior are an important source of fun in game playing. AI exists in almost all game designs. However, through the interactive stimulus between progressive software and hardware in game field, the game pictures evolve from 2D plane in the early time, 3D stereo in the mid time, to high resolution today. The showing effects of pictures in game constantly shock the players. However, comparing with large progress in picture showing of games, the development of AI seems much slower. Regarding this, this research attempts to study and develop the AI system of an Action Game based on the platform of XNA Game Studio. Here we totally make three application cases to illustrate the AI effects in game design. In our proposed AI system, object-oriented modulated concept is introduced, and every game function and AI technique is set as independent event. Through operations of combination and permutation of proper events, the AI system may possess the properties of usage repeats, design elasticity, and design enlargement. Meanwhile, techniques of finite-state machine, fuzzy-state machine, and artificial neural network are also used in AI system. These techniques are then combined with rule-based and adaptive learning systems to promote the learning ability of roles with wisdom in game playing so as to enhance their human imitation and funs for players.