A method and system for identifying and acting on a handwritten action item is disclosed. The system may learn a set of user-defined symbols and associate each symbol with an action category. Then, when the system that captures a handwritten action item that includes one of the symbols, it will determine which action category that corresponds to the symbol, identify process parameters in the action item, determine a task to be performed based on the action category, and apply the process parameters to automatically perform the task.