Agile development methodologies are becoming increasingly popular because they address many risks of software development through things such as quick delivery of working software and responsiveness to change. As agile organizations have begun to develop more user interface-intensive systems, they understand the value and need to design more usable systems. The fields of usability engineering and human-computer interaction are focused on exploring how people interact with computer systems. However, much of this work is inaccessible to agile practitioners because it does not align with core agile values and because there has not been adequate transfer of knowledge between practice and academia. This motivated my creation of the eXtreme Scenario-Based Design (XSBD) process, an integrated agile usability approach. XSBD provides key usability benefits of the scenario-based design (SBD) approach (an established usability engineering process) and is compatible with an agile development framework modeled on leading agile processes like XP and Scrum. XSBD was designed for use in projects in which a large part of the overall system quality is determined by system usability. This requires close communication and coordination of the disparate usability and agile development work practices. A core aspect of XSBD is the central design record (CDR), which is the shared design representation that guides usability design. It tightly couples usability evaluation results to design features and high level project goals, allowing the usability engineer to leverage key benefits of traditional SBD while working in an agile framework. I began developing XSBD at Virginia Tech, evaluating it through several student-led development efforts. To improve and demonstrate the applicability of XSBD in practice, I partnered with Meridium, Inc., a software and services company. Using an action research case study method, I worked with several development teams there who used XSBD to develop products. This directly linked usability and HCI research to practice, allowing me to demonstrate XSBD's utility in practice while evaluating it from a theoretical perspective. The results of this work suggest several avenues for further work both to increase its adoption in practice and to link to existing HCI research efforts such as design rationale and knowledge reuse. / Ph. D.