Operationalizing and Optimizing Human–Agent Teaming for Knowledge Work

Research suggests that artificial intelligence (AI) will not completely replace human knowledge workers in the near-term future (Woodruff et al., 2024; World Economic Forum, 2025). Rather, recent AI developments will support and augment human capabilities. Knowledge work involves dynamically novel, collaborative tasks that require workers to make sense of large amounts of new information, to draw on deep domain expertise, and to continually update their situational awareness (of, say, a complex adversary or business competitor, its goals, and its actions), while accounting for unavoidably incomplete knowledge due to information overload and potentially inaccurate assumptions about the quality of one another's contributions (Kane et al., 2018, 2023, 2025). Given the complexity and open-ended nature of knowledge work, AI will be most effective when providing critical input to workers, rather than by trying to independently “solve” the key problem, whether that be determining the risk of an adversary's troop movement or the wisdom of opening a new business venture. Thus, the fundamental challenge we explore in this contribution is understanding how human–agent teams (HATs) performing knowledge work can most effectively structure their work. This chapter presents our conceptual framework for HAT orchestration and motivates the creation of testbeds to facilitate in the operationalization of this framework.