We present in this position paper the foundations of the MIC* model and deployment environment as they have emerged in the last years, relate them to current evolutions on semantic GRID dynamic service generation, as reflected by OGSA/I and, more recently, by WSRF, and propose an integrated view with the previously proposed STROBE model for communicating agents leading to a quite simple yet very promising architecture that may include Human agents in the loop, uninspectable as most artificial agents, yet behaving autonomously by interacting. Modeling Human agents seems to us of high importance in the high level service generation required by complex, semantically rich applications, such as those of e-Learning, e-Science or e-Commerce envisaged in the years to come.