Industry is looking for practical means to improve the accuracy of the parts machined on CNC machines. Some artificial intelligence (AI) systems have been applied in modeling and compensating manufacturing process errors in CNC machining. However, these systems are not capable of predicting the results of a new operation if no sufficient data on a number of similar operations is available. A generalized AI approach named synergistic interactions amongst modeling, sensing and learning is proposed in this paper. Based on the AI approach, a new strategy of error compensation of workpiece dimension in CNC machining is developed and applied in a CNC turning center. Error compensation results are illustrated the effectiveness of the error compensation strategy. The learning curve shows that the error compensation confidence gradually progresses towards 100% confidence from zero along with the CNC machine operation time increasing.