A complex system is usually hard to describe, because it consists of several elements that interact with each other. Those interactions are often, and hence not directly proportional to their causes. Such complexity demands new approaches to design, control and optimize. This work explore the use of the COllective INtelligence (COIN) approach to address the challenges of the complex systems. A Collective is as a multi-agent system where each agent is self-interested, capable of learning. Also, the system has a well deﬁned objective function that rates the performance of the Collective. To demonstrate and explore the potential of the COIN theory, three experiments were investigated. In the ﬁrst experiment, we investigated the power of the COIN approach to optimize a variant of the congestion game with full communication level. In the second experiment, we investigated how the COIN approach behaves when there is a communication restriction among the agents. Finally, we applied the COIN theory to the network packet routing problem with diﬀerent topologies.