Developers of future systems are relying on graphical user interfaces (GUI) for user interaction. As the interfaces get more powerful and complex, it becomes increasingly important to be able to predict how well they will work prior to absorbing the cost of implementation. This paper describes the development, application, and future plans for the Graph-Based Interface Language (GRBIL) tool. GRBIL can be used to predict human performance on new GUI designs, resulting in quantitative comparisons between interface alternatives. This enables human factors engineers to participate in system design processes and achieve influence very early. Using GRBIL, an interface designer can sketch the proposed interfaces, describe the tasks that the user will perform, and even inject stressor effects to represent fatigued or under-trained users. GRBIL will generate estimates for the times to perform the tasks and will predict whether errors (such as those caused by memory lapses) are likely. The designer can then use GRBIL to adjust the GUI to explore alternatives that might improve performance.