<p>Knowledge graph completion (KGC) is crucial for enhancing the predictive capabilities of knowledge graphs (KGs) by inferring missing links and entities. In such domains as telecommunications, where business processes are highly interconnected, an effective KGC model can serve as a recommendation system for business analysts, aiding in process optimization. This paper proposes a transformer-based multi-hop reasoning model for KGC, trained on a telecommunications KG structured according to the models provided by TM Forum consortium that is a global industry association for service providers and their suppliers in the telecommunications industry. The model leverages relational paths and semantic dependencies within a sequence-to-sequence architecture to predict new connections and generate interpretable reasoning paths. To address the scarcity of real-world data, a novel method for generating synthetic telecommunications KGs with controlled structural variations is introduced. Experimental results demonstrate the model’s robustness and superior performance compared to traditional embedding-based methods, particularly in noisy environments, while providing actionable insights for business analysts. The work bridges the gap between KGC and practical recommendation systems, highlighting the value of interpretable multi-hop reasoning in telecommunications business process optimization.</p>