The emergence and development of the complex adaptive systems (CAS), such as industrial clusters, call for a radically different managerial approaches and leadership strategies than in the classical stable hierarchical systems. Their leaders have to take into account the key attributes of CAS, such as self-organization, inter-relatedness, adaptiveness and emergence. The complex adaptive leaders have to address the contradictory qualities of networks and tensions stemming from the dynamic interactions of their actors. This paper aims to present and discuss the different strategic approaches to the development of industrial clusters as CAS, based on the findings of research carried out in nine Lithuanian clusters representing a variety of industrial settings. The semi-structured interviews with cluster leaders as "boundary spanners" of networks enabled to distinguish 3 different strategic approaches at 2 different stages of cluster development. At the initial stage of cluster legitimation, the experimentation strategy, external agent strategy and strategy of ad hoc clustering of fellow professionals can be distinguished. At the later stages, strategies vary by the degree of involvement from strategy of self-organisation and strategy of coordinated self-organisation to strategy of active coordination. Industrial clusters represent a specific type of complex adaptive systems, where the role of leadership is not confined to the central figure of network coordinator, but is dispersed across many agents of the system. The proactive leadership strategies adopted by cluster coordinator (as change agent) are particularly important at the initial stages of cluster formation. The more advanced and more the relationships, the more oblique is the role of facilitator. The leaders should therefore understand the continuously evolving nature of cluster, the importance of balancing the self-organization with active coordination approach, the sense of common purpose with individual interests and objectives. The adopted leadership strategies must be dynamic and flexible enough to accommodate the evolving nature of the specific cluster environment.