Self-organizing emergent systems, also referred to as Decentralized Autonomic Computing systems, are commonly known for their scalability, robustness, flexibility, and adaptivity rather than their efficiency. However, certain application scenarios, in particular in industrial settings, require a high degree of efficiency from these systems as well, in order to keep operational expenditures and energy use small. In this paper, we therefore present the concept of an advisor, designed to improve the efficiency of self-organizing emergent multi-agent systems solving industrial problems with recurring tasks. The advisor autonomously identifies the recurring tasks at runtime and provides the agents with advice for better solutions in the future, if indicated. The advisor does not limit the self-organizing behavior of the underlying system, i.e., all problem-solving decisions are still locally made by the agents. Experiments with instances of dynamic pickup and delivery problems show that the advisor concept can achieve substantial efficiency improvements, even if the recurring tasks change over time.