The primary point of business is the offering stage, which is crucial for success. Property marketing during a pandemic presents unique challenges that require strategies to minimize the impact. To determine effective marketing policies and strategies, it is necessary to predict and understand customer desires. Predicting customer behavior when making purchases can have significant implications, including providing more targeted and efficient advertising to increase sales. This study aims to analyze the ability to predict prospective customers' purchasing behavior using deep learning and pre-trained models. The data for this analysis was collected through follow-up with prospective customers by sales personnel and processed using techniques such as text embedding with FastText and BERT, followed by training with FastText, BERT, and BiLSTM. The results showed that the FastText model achieved the highest performance, with an accuracy value of 94.19%, an AUC value of 92.03%, and 94.28% precision. The model was then implemented into a web application using the Flask framework.