Abstract
Agentic Artificial Intelligence (AI) extends delegation beyond task-level hand-offs to process-level execution. While widely promoted as a pathway to efficiency, the assumption that “more delegation means more efficiency” remains untested. This study examines how two agentic AI design strategies – multi-agent versus single-agent – shape efficiency outcomes in IT support. We conduct a three-phase mixed-methods study involving: (1) a qualitative exploratory phase capturing employees’ experiences of multi-agent and expectations of a single-agent design, (2) a field experiment comparing resolution time and cost-adjusted human effort, and (3) a qualitative explanatory phase to interpret outcomes. Preliminary findings suggest that multi-agent design enhanced efficiency by redistributing repetitive tasks and preserving escalation structures, while single-agent design was perceived as both promising greater efficiency and risking reduced oversight. Upon completion, this study will provide evidence on whether delegation inherently produces efficiency and develop design knowledge on how delegation can be configured to achieve organisational outcomes.
Recommended Citation
Batra, Yagyaa; Tim, Yenni; and Kirshner, Sam, "Does More Automation Mean More Efficiency? A Mixed Methods
Study Exploring Agentic AI Design Strategies" (2025). ACIS 2025 Proceedings. 75.
https://aisel.aisnet.org/acis2025/75