The leadership question has changed. Five years ago, executives asked: should we adopt AI? Today the question is: how do we lead teams where humans and machines work together, where algorithmic systems make recommendations that humans may override or follow, and where the boundary between human judgment and machine output is blurred? This shift from adoption to orchestration requires a new leadership competency that current management frameworks do not adequately address.
Trust and Weight
Wen et al. (2025), in Frontiers in Organizational Psychology, investigate the dynamics of trust in human-AI collaboration within organizational management. Their research examines how leaders and team members calibrate the weight they give to AI recommendations versus human judgment. The findings reveal that trust in AI is not a single construct but varies by task type, perceived stakes, and the individual's prior experience with AI systems.
High-performing teams develop what the authors call calibrated trust — an accurate assessment of when AI recommendations are likely to be reliable and when human judgment should take precedence. Leaders play a critical role in developing this calibrated trust, not by mandating AI adoption or allowing uncritical AI dependence, but by creating environments where team members learn through experience which tasks benefit from AI input and which require human override.
The organizational design implications are significant. Traditional authority structures assume that decision rights flow from position in a hierarchy. AI-augmented teams require decision rights that flow from competency — sometimes the AI has better judgment (data-rich, pattern-recognition tasks), sometimes the human does (novel situations, ethical considerations, stakeholder relationships). Leaders must design decision architectures that route decisions to the most competent agent, whether human or machine.
Redesigning Authority
Westover (2025), in the HCL Review, proposes a framework for AI-augmented decision rights that explicitly redesigns how authority operates in organizations where AI plays a substantive role in decision-making. The framework distinguishes between decisions that should be automated (high-volume, well-defined, data-rich), decisions that should be AI-assisted (complex but with available precedent), and decisions that should remain human-exclusive (novel, values-laden, politically sensitive).
The framework's contribution is practical: it provides leaders with a structured approach to the delegation question that AI introduces. Traditional delegation theory assumes delegation to another human — someone who can be held accountable, who can exercise judgment, and who can explain their reasoning. Delegation to AI requires different accountability structures, different feedback mechanisms, and different forms of explanation. Leaders who treat AI as just another team member — or who refuse to delegate to AI at all — both miss the productive middle ground.
When Automation Fails
Krzywdzinski et al. (2025), in AI & Society, examine what happens when AI-augmented systems fail — and how team organization influences the ability to diagnose and resolve automation failures. Their research finds that teams with rigid role definitions and limited cross-functional knowledge struggle more with automation failures than teams with flexible roles and shared understanding of how automated systems work.
The implication for leadership is that AI-augmented teams require investment in human capability, not just AI capability. Team members need sufficient understanding of AI systems to recognize when they are failing, diagnose why, and implement manual workarounds. Leaders who invest only in AI deployment without investing in human capacity to manage AI failures create teams that are more productive under normal conditions but more fragile under stress.
This fragility problem is the central tension of AI-augmented leadership. AI increases average performance while potentially increasing variance under failure conditions. The leader's role is to capture the performance gains while building the resilience that prevents catastrophic failures — a balancing act that requires understanding both the capabilities and the limitations of every agent on the team, human and machine alike.
The Fragility-Performance Trade-off
The central tension of AI-augmented leadership deserves deeper examination. AI increases average team performance by automating routine analysis, accelerating information processing, and eliminating certain categories of human error. But it also introduces new fragility — the team becomes dependent on systems that can fail in ways that team members may not understand, at times that are difficult to predict, and with consequences that cascade through interconnected workflows.
Managing this trade-off requires what might be called operational metacognition — organizational awareness of its own dependence on AI systems and deliberate investment in the human capabilities needed to function when those systems fail. This includes maintaining manual backup processes for critical functions, cross-training team members to understand AI system operations, and regularly exercising failure scenarios so that teams develop muscle memory for operating without AI assistance.
The leadership challenge is that investments in resilience are invisible when AI systems work correctly. The team that maintains manual backup capabilities looks less efficient than the team that has fully automated — until the automation fails. Leaders who invest in resilience must justify invisible insurance against events that may never occur, while leaders who maximize AI-driven efficiency receive immediate credit for measurable productivity gains. This incentive asymmetry systematically under-invests in resilience unless leadership explicitly values and rewards it.