The study presents a new approach to diagnosing Alzheimer's disease that combines two types of artificial intelligence. The first model (ANN) processes clinical data from 1,200 patients based on 31 different features, including demographics, symptoms and patient behavior, and achieved an accuracy of 87.08% in predicting risk at an early stage of the disease. The second model (CNN) analyzes brain images from MRI and determines the stage of the disease with 97% accuracy. The combined two-model system effectively combines structured clinical information with brain image analysis, thereby overcoming the limitations of traditional diagnostic methods. Grad-CAM visualizations improve understanding of how the model makes decisions, which supports its use in clinical practice. This approach provides a more comprehensive assessment of Alzheimer's disease and helps clinicians make decisions. Future developments will focus on simpler versions of CNNs and wearable technologies that would enable wider availability and earlier intervention.