Graph neural networks (GNNs) have emerged as powerful tools in the realm of graph machine learning, particularly in applications like building recommendation systems. In this context, Memgraph has developed a new module leveraging GNNs for creating a recommendation system tailored to telecommunication packages. GNNs, distinct from convolutional neural networks, are adept at capturing complex relationships in data by iteratively aggregating node representations from their neighbors. This capability is harnessed using algorithms like GraphSAGE and Graph Attention Networks (GAT), which effectively scale and prioritize information within large graphs. The process involves constructing a graph from a dataset of customer-service interactions, with additional connections based on shared demographics to enhance learning. Despite the challenges of training such models, including hyperparameter tuning, the system demonstrates promising performance metrics like accuracy and F1 scores around 90%. Utilizing tools like Memgraph and DGL, the system predicts links between nodes, offering recommendations with metrics such as Precision@k and Recall@k, showcasing the potential for GNNs in practical applications.