A system processes a set of instructions, each of which indicates an action to perform on a user interface. The system does not have prior information concerning the layout of the user interface, nor does the system interact with the user interface through an automation API. For each instruction in the set, the system automatically performs the indicated action on the user interface. To do so, the system identifies the selected control on an active window, and determines whether it is the control to activate in order to perform the indicated action. If so, the system generates an input signal configured to activate the selected control, and sends the input signal to the user interface. If not, the system selects a new control and determines whether it is the desired one. The system cycles through the controls to find the desired one to activate.