An image of a graphical user interface is captured. A process is started to record identification of a graphical object (e.g., a text field) in the image of the graphical user interface. Recording identification of the graphical object in the image of the graphical user interface is based on machine learning. An end to the process to record identification of the graphical object in the image of the graphical user interface is identified. A machine learned software object is updated. For example, the machine learned software object is updated with a location of the graphical object. The updated machine learned software object then is used to test the graphical object. For example, to automatically fill in the text field using a test script.