In a multi-agent dynamic system, the optimal actions by each individual agents (based on whatever cost-benefit analysis they do) can evolve the system into a state where every agent is worse off compared to some initial state. This holds even if every individual agent is a "smart businessperson".

One main purpose of law and social rules is to prevent multi-agent systems from getting stuck into these global non-optimal states. And arguing that agents are smart is not a counter-argument to this.