Robotic Process Automation (RPA) has become a promising approach to automate repetitive processes by mimicking the way in which humans interact with user interfaces (UIs). However, RPA robots may fail to do so due to unexpected changes in the UIs, which alter the UI interaction flow and can result in exceptions being raised. This can create the need for human intervention, which can be time and resource costly. To these issues, existing work has identified the need for automatic exception handling and the addition of cognitive technologies within RPA processes. To address this problem, we propose a hybrid approach that aims to integrate AI agents —more specifically, Graphical User Interface (GUI) agents— in the RPA lifecycle. More precisely, we propose to integrate them within the monitoring stage —when the automation is already deployed— with the objective of handling UI-related exceptions by taking control of the process execution when needed. Our findings show promising results for this approach to be integrated as a solution to UI-related exceptions during robot execution; thus, potentially reducing the need for manual intervention and preventing automation workflows from being blocked until robots are updated.