As 6G architectures evolve toward pervasive Network Intelligence (NI), managing concurrent and potentially conflicting control actions from diverse Artificial Intelligence (AI)-driven xApps becomes a critical challenge. In the disaggregated Open Radio Access Network (O-RAN) ecosystem, independent xApps optimizing for different metrics, such as energy efficiency versus throughput, can lead to network instability and degraded performance. This paper presents a conflict management solution within the ORIGAMI framework, specifically addressing the contention between the Interoperable Machine Learning models for RAN Energy efficiency (IMLE) xApp (energy optimization) and a competing Throughput xApp (throughput maximization). We demonstrate how ORIGAMI's Global Service-Based Architecture (GSBA) facilitates the real-time discovery and mediation of these conflicting objectives. Our experimental results in a distributed O-RAN testbed show that by leveraging GSBA-exposed services, the proposed conflict management logic effectively prioritizes critical network requirements, ensuring stability and resource efficiency without compromising the performance targets of high-priority services.