In this paper, my coauthors and I examine how organizations can design more human-centred systems as AI becomes embedded in core management functions such as hiring, training, performance evaluation, and compensation. We show that the key design questions (e.g., when humans and AI should share decisions) depend not only on how routine or cognitively complex a task is, but also on whether people see AI’s role as acceptable and fair—especially in high-stakes decisions that affect careers and livelihoods. The broader message is that designing human-centred organizations means building systems that improve efficiency without losing fairness, accountability, and the vital role of human professionals in shaping important decisions.

This paper explores what it means to build organizations for a world where social interaction increasingly takes place in immersive virtual environments. We argue that the metaverse is not just a technology platform, but a new organizational setting that could reshape how people collaborate and assume social roles through digital representations such as avatars. Instead of asking whether virtual worlds will replace face-to-face interaction, we explore how organizations can use them to support more engaging forms of connection than typical remote-work tools allow. The takeaway is that designing human-centred organizations for the future means thinking seriously about which forms of connection, trust, and collective work can be re-created in new digital environments.

Here, I explore how work itself may need to be redesigned as people increasingly collaborate through multiple AI agents acting on their behalf. I propose a new model in which each person leads a team of AI representatives with different roles, levels of authority, and degrees of closeness to the human’s current goals and preferences. I emphasize that organizations must put strong safeguards in place so people know when they are interacting with AI, humans remain responsible for critical decisions, and technology strengthens rather than erodes human connection. In short, this paper asks how we can redesign workflows around distinctly human strengths while letting AI take on more of the coordination burden.

Multi-agent AI systems are often described in technical terms, including, communication protocols, reward functions, and reasoning architectures. This article explains why this framing is incomplete. A century of research in organization science has identified four universal problems that any multi-agent goal-oriented system must solve. Multi-agent AI systems instantiate these problems by construction. The problems are old; new solutions remain to be discovered.