Alzheimer’s Disease AD is a neurodegenerative disease that gets bad with time and it has no known cure. The research focuses on early detection, individualized caregiving, and an all-encompassing approach to healthcare. We present a comprehensive approach to early AD detection through the development and implementation of a Convnet model, which is enhanced by using transfer learning techniques including (Dense121, Restnet, and InceptionV3). The proposed model performs four-class image classification to distinguish between Mild, Moderate, Very Mild, and Non-Demented states. Capsule Networks (CapsNet) architecture integrated, which accelerates in preserving spatial sequential hierarchy in image data Explainable AI (XAI) techniques have been incorporated, particularly, LIME and LRP are used to influence how neural networks make decisions, ensuring that classifications are transparently and comprehensibly justified. This makes it easier to integrate neural networks into clinical practice.