A system and method are disclosed that trigger a smart response to a button action when two or more running applications have applicable actions on a mobile device. The system includes a machine learning algorithm (MLA) built into the operating system that analyzes and learns from user actions. When the system detects a hardware button press, it retrieves possible actions/events that may be triggered for the current app and other running apps and processes. The actions are evaluated by computing a combination of scores involving machine-learned, rule-based scores provided by the apps and real-time signals such as location, time, user activity, etc. The system then triggers an action that has the highest score surpassing a predetermined threshold. The disclosed method would provide improved user experience, and also new, useful button actions that previously were not available to the user.