This paper introduces a novel method for engineering trustworthy artificial intelligence (AI) by integrating the agile Ethical, Legal, Societal, and Teaming Assessment Framework (ELSTAF). The ELSTAF is a process for identifying human-centered design and deployment priorities for agentic AI and measuring performance against sociotechnical design requirements generated throughout the engineering lifecycle. We are piloting this framework alongside our colleagues who are developing a set of semi-autonomous decision support agents for command, control, and communications battle management (operational) operators within the United States Department of Defense. This paper documents the procedural choices of our interdisciplinary research team and summarizes preliminary findings from the first implementation of our proposed framework. As our research progresses, we will complete the final two phases of the ELSTAF framework following the methodology described, to assess agent performance and sociotechnical alignment within increasingly high-stakes, uncertain, and error-prone situations. This research will offer valuable insights into the future of agentic AI development for decision support systems in complex operational environments.