Convolutional Neural Networks (CNNs) have significantly impacted various fields, including image classification and natural language processing, by delivering state-of-the-art performance. Despite their success, CNNs are often criticized for their black-box nature, where the reasoning behind their predictions remains opaque and difficult to interpret. This lack of transparency raises concerns, especially in critical applications such as healthcare and autonomous systems, where understanding model decisions is crucial for trust, accountability, and regulatory compliance. This work aims to address the interpretability challenges associated with CNNs by integrating Explainable AI (XAI) techniques with the principles of white-box AI. We propose a novel framework that combines post-hoc XAI methods, such as Gradient-weighted Class Activation Mapping (Grad-CAM), Layer-wise Relevance Propagation (LRP), and SHapley Additive exPlanations (SHAP), with inherently interpretable model architectures, including Self-Explaining Neural Networks (SENN). The proposed approach is designed to improve the transparency of CNNs without sacrificing their performance. Through extensive experimentation on benchmark datasets such as CIFAR-10 and ImageNet, we evaluate the effectiveness of our framework in terms of accuracy, fidelity, coherence, and user trust. The results show that our architecture enhances interpretability while maintaining competitive accuracy, making it a viable solution for high-stakes applications requiring both precision and explanation. By advancing the understanding and transparency of CNNs, this work contributes to the development of AI models that are not only accurate but also explainable and trustworthy, paving the way for safer and more reliable AI systems.