Traditional credit risk models, which rely primarily on static and historical financial records, are increasingly insufficient in addressing the complexities of modern economies. Their retrospective orientation often fails to capture the real-time operational health of borrowers, resulting in suboptimal lending decisions. This study proposes a novel smart framework that integrates high-frequency Internet of Things (IoT) data streams with Explainable Artificial Intelligence (XAI) methods to enable dynamic and transparent credit risk assessment. The architecture incorporates diverse real-time operational signals—including supply chain logistics, equipment condition, production volumes, and inventory status—to construct continuously updated borrower risk profiles. At its core, the framework combines a Graph Neural Network (GNN) to capture intricate interdependencies within supply chains with a Long Short-Term Memory (LSTM) network for temporal analysis of IoT sensor data. An additional XAI layer, implemented through SHapley Additive exPlanations (SHAP), ensures interpretability of model outputs, thereby supporting regulatory compliance and fostering stakeholder trust. To evaluate the framework, a hybrid dataset was constructed, combining traditional financial statements with simulated IoT streams that mimic realistic business operations. Experimental results highlight a substantial performance improvement over conventional approaches, achieving an Area Under the Curve (AUC) of 0.97. Moreover, the XAI module generated transparent, feature-based explanations for changes in risk scores, offering actionable insights for lenders. This research argues that the convergence of IoT and XAI signals a paradigm shift from static, retrospective risk models to proactive, dynamic, and interpretable credit risk management, enabling financial institutions to make better-informed and timely lending decisions.