Recent developments in IoT, cloud computing, and AI have revolutionised the conventional healthcare system and set the stage for the rise of smart healthcare. Significant advancements in healthcare are possible through the combination of Internet of Things and artificial intelligence. Opportunities abound in healthcare thanks to the combination of IoT and AI. In light of this, this study presents a new approach to disease diagnosis using the intersection of AI and IoT in the context of intelligent healthcare delivery systems. The primary goal of this article is to use AI and IoT convergence techniques to create a disease diagnosis model for heart disease and diabetes. Data collection, preprocessing, classification, and fine-tuning are all parts of the proposed model. Wearables and other sensor-based IoT devices facilitate effortless data collection, which is then utilised by AI methods for disease diagnosis. Combining the Crowd Search Optimisation (CSO) algorithm with the Cascaded Long Short Term Memory (CLSTM) model, this study presents a novel approach to disease diagnosis. Applying CSO to the CLSTM model's 'weights' and 'bias' parameters improves its ability to classify health data. The isolation Forest (iForest) method is also used in the study to filter out anomalies. When CSO is incorporated into the CLSTM model, diagnostic accuracy is greatly improved. Extensive experiments were performed using healthcare data to validate the performance of the CSO-CLSTM model. The outcomes showed that the proposed model performed exceptionally well, with an accuracy of 96.16 percent for detecting heart disease and 97.26 percent for detecting diabetes. Therefore, the CSO-CLSTM model emerges as a robust and efficient instrument for disease diagnosis, amenable to incorporation in intelligent healthcare systems.