The Internet of Things (IoT) is an emerging field for researchers. It is a network of public objects that are embedded with technologies that help them communicate and interact within themselves and the external environment, for example, a smart home, and its goal is to provide people with services that suit their needs. As the traditional process of preparing Arabic coffee, this paper aims to improve the process of preparing Arabic coffee using machine learning and computer vision techniques, by addressing the problem of bubbling that occurs during traditional preparation, which requires careful supervision. A comprehensive dataset of 70 videos of the coffee preparation process was collected, resulting in the extraction of 834 images that were divided into training and validation sets. The pre-preparation steps included tagging objects using the Roboflow tool and rotating the images, which increased the size of the dataset by four times. Using the YOLOv8 algorithm, a model was trained to detect and track the state of coffee in real time. Before data augmentation, the model achieved an mAP50–95 of 0.636. After data augmentation, these metrics improved significantly, reaching an mAP50–95 of 0.917. The trained model was tested on live video streams and successfully monitored the coffee status in real time and alerted users when brewing was complete, preventing spills. The system demonstrated high efficiency and accuracy in performance, mimicking human coffee brewing methods. The results demonstrate the potential of the YOLOv8 algorithm to automate and optimize routine tasks with high accuracy.