The contribution reports on three aspects of our research activities on GRID services, Agents and Human learning: an integrated vision, a statement of intentions concerning a relatively new life cycle for Service Engineering and a review of achieved results, presented by embedding remarks and quotations in the relevant points. The essence of the contribution lies in the concept of service that is considered to be intrinsically conversational both during its dynamic definition and during its delivery. It is shown that Agents are the most promising abstractions (and technologies) offering a concrete approximation for future conversational GRID services and that Human learning is a quite suitable context for including the Human in the loop of the higher level services to be developed for mixed Virtual Organizations on future GRID networks.