Accurate credit risk assessment for Small and Medium-sized Enterprises (SMEs) is crucial for financial institutions, yet traditional methods often fail to capture the complex interconnections within the SME ecosystem. However, working with SME data presents significant challenges, including the limited availability of high-quality data, inconsistencies in director records, and variability in data formats. In this study, we address these challenges by leveraging graph-based approaches to model the intricate relationships between SMEs and their directors, focusing on how these relationships impact SMEs’ liquidation likelihood. By representing SMEs as nodes and shared directors as edges, we adopt the “Entity-Attribute Relationship” paradigm to construct a graph that encapsulates both individual and relational attributes. Using a Graph Convolutional Network (GCN), we evaluate the predictive power of graph-based methods compared to traditional logistic regression. Our results demonstrate the superiority of GCN over the logistic regression baseline as an industry standard. This highlights the value of incorporating graph-structured data, such as shared director relationships, into credit risk modeling. This research contributes to the growing body of literature on AI-driven credit risk assessment, particularly for SMEs, by showcasing the effectiveness of graph-based features in capturing critical relational dynamics while addressing the inherent challenges of working with SME data.