The production and ramp-up of complex and highly customized products are exceptionally challenging for planning and control, especially in small lot sizes. Daily challenges like late requests for change, immature high technology products and processes create significant risks. The occurring risks are bigger in big series production, such as automotive. Thus, new ICT-based approaches are required. The aim is to develop mitigation strategies to respond faster to unexpected events. Therefore the knowledge base has to be enriched for real-time decision support, to detect early warning and to accelerate learning. Our approach is based on a new generation of service orientated enterprise information platforms, a service orientated bus integrating service-based architecture and knowledge-based multi-agent systems (MAS). A holonic MAS combined with a service architecture will improve performance and scalability beyond the state of the art. The solution integrates multiple layers of sensors, legacy systems and agent-based tools for beneficial services like learning, quality, risk and cost management. Additionally, the ecological footprints will be reduced. The ARUM solution will run in two modes: predictive and real-time simulation. The predictive mode supports the planning phase, whereas the real-time operations mode supports dynamic, time-, cost- and risk-oriented re-planning of operations. In case of immaturity or late requests for changes, alterations of engineering information provided is supported equally. ARUM is strongly end-user driven and the results will be tested on three industrial use cases with focus on aircraft, aircraft interiors and ship manufacturing. The solution will be validated in a real industrial environment by industrial partners and benchmarked against today’s ICT solutions. In collaboration with universities, a test-bed will be established for ARUM systems and tools designing and testing and will be open for dissemination and demonstration.