Technology and society are in an interminable process of mutual effect on each-others transformation; thus, computational challenges along evolving social dynamics are getting highly complex. New approaches are needed to tackle this complexity, as some computational tasks cannot be solved solely by software. Consequently, we argue that systems need to be evolving toward the integration of software services, things and people that work together, e.g., as Collective Adaptive Systems. This thesis focuses on the people part of these systems, and more precisely on automating the provisioning and management of human-based resources, in scenarios where applications and systems include human-computation. Human computation has been slowly but firmly gaining its momentum, and software is beginning to be designed and built with the possibility of enabling human task-execution to be provisioned as a service. In this work, we investigate the mechanisms to provision and elastically manage collective and collaborative human computation. We present a report of our investigation of strategies for formation and elastic coordination and management of collectives of experts who provide their skills online as services. With particular focus on elasticity we argue that human computation is more efficient, reliable, on-time and cost-effective when expertise can be scaled in and out, at runtime. Moreover, we argue that trust is highly important due to the many uncertainties that come from the human nature. Hence, we investigated trust metrics and ways of using trust in automated coordination of collectives. Furthermore, we see runtime negotiations as an additional mechanism to guarantee quality of results for online human-task execution and argue that they are as crucial for human-based services as they are for software services. Last but not the least, we investigated the benefits of storing provenance data, along with the challenges that human-computation entails regarding privacy.