The Automatic Generation Control (AGC) system provides secondary frequency regulation by commanding online generating units in order to compensate the active power imbalances in power systems. System operators are responsible for monitoring and evaluating the performance of this service that guarantee the secure operation. However, due to the complexity of the AGC system and the amount of available data, the performance evaluation could be challenging. As an alternative, a data-driven approach was used to develop a tool that support the operator performing this task, providing an insight in a few minutes. The proposed machine learning-based tool for AGC performance monitoring and evaluation was tested on the Colombian system operator.