Grid computing offers a service for sharing resources over uncertain and complex environments. In such multi-participated settings it is essential to make the grid middleware functionality transparent to members by providing the ability to act autonomous and learning from the environment. Parallel to grid, artificial neural networks is a paradigm for processing information, which is inspired by the processes of biological nervous systems. The latter fields can be really promoted from the artificial intelligent agents, which offer an autonomous acting infrastructure of members with proactive and reactive aptitude. As grid is about sharing and allocating resources within uncertain domains, intelligent agents and neural networks may be the mean of achieving an autonomous learning environment of self-motivated members. In this study, we focus on the mobility agents' model aiming to discovery resources dynamically, parallel to the artificial neural networks as a way to achieve the best resource discovery paths. Our work is fundamentally based on the Self-led Critical Friends method, a technique for realizing inter-cooperation among various scales Virtual Organisations (VOs). Their mediator acting nature redirects communication to other parties of different VOs by utilizing a public profile of data stored within VO members.