Estimating the application performance in distributed scenarios is challenging due to numerous variables and influences. Combining Artificial Intelligence (AI) with advanced monitoring methods offers a promising approach for predicting application quality. State-of-the-art solutions leverage robust monitoring and AI algorithms, such as Deep Neural Networks (DNNs), to detect temporal relationships and improve estimation accuracy. However, these methods often require significant computational resources and impose overheads due to the fine-grained monitoring requirements. This paper proposes an alternative hypothesis: effective insight into the application’s behavior and user perception does not rely solely on highly granular metrics. Instead, lightweight AI algorithms combined with generic infrastructure metrics–such as computing, networking, storage, and operating system variables–can yield valuable patterns. Our method achieved lower Mean Absolute Error (MAE) and Mean Squared Error (MSE) in forecasting write-and-read operation latencies in a database cluster, demonstrating its effectiveness and efficiency.