We present a study of the information transferred among caregivers in the context of cardiac surgery and use the study to evaluate a system, MAGIC, that we are developing for automated generation of briefings. Our framework integrates cognitive and quantitative evaluation methods and features three standards that reflect current practice in the Cardiothoracic Intensive Care Unit (CTICU). Using experimental design to compare human-generated and machine-generated briefings, we show that MAGIC's current level of performance is useful. Moreover, MAGIC could help improve information flow in the CTICU by providing a consistent set of information earlier than in current practice. The separate standards are also consistent in suggesting specific modifications that may be necessary for iterative design and further system development.