The Orchestration Imperative: Six Capabilities for Leading Human-AI Teams

The integration of agentic AI into organisational workflows represents a structural shift in how work is conceived, distributed, and governed. As intelligent systems move beyond narrow automation to assume responsibility for complex, multi-step tasks, the human role is transitioning from direct execution to orchestration, review, and strategic direction (Mollick, 2024). This transition demands capabilities that most organisations have neither identified in their workforce planning nor developed through their learning strategies.

Research from Harvard Business School provides a useful illustration. When knowledge workers collaborate effectively with AI, productivity gains of up to 40 per cent are achievable. The same research reveals that workers who lack the judgement to recognise the boundaries of AI capability can produce work of significantly lower quality than those working without AI at all (Dell'Acqua et al., 2023). The value of human-AI collaboration depends entirely on the orchestration capability of the human.

From operator to orchestrator

For most of the past century, organisational design has assumed that the primary unit of work is the task, and that the primary role of the worker is to execute that task. Training systems, performance management, and career progression have all been built on this assumption. The arrival of AI that can execute complex, multi-step tasks autonomously breaks this model at its foundations (Susskind, 2020).

Agrawal, Gans and Goldfarb (2022) describe this as a shift from "prediction machines" to "decision systems." When AI provides the prediction, the human role is to provide the judgement. When AI handles the execution, the human role is to provide the direction, the review, and the accountability. This requires a fundamentally different capability profile from the one most organisations have invested in developing.

Daugherty and Wilson (2018) identify three categories of new human role that emerge as AI takes on execution: trainers, who teach AI systems how to perform; explainers, who interpret AI outputs for stakeholders; and sustainers, who ensure AI systems operate responsibly over time. We extend this taxonomy to include a fourth role that has become central in the agentic era: the orchestrator, who coordinates human and AI contributions across complex workflows and ensures the combined output is greater than either could achieve alone.

The six capability domains

Drawing on research from organisational psychology, technology management, cognitive science, and AI ethics, we have identified six interconnected capability domains that together constitute the orchestration skill set. Each domain is a necessary condition for effective human-AI teaming. None is sufficient on its own.

Domain 1: AI Fluency and System Interaction. The foundational ability to communicate with, prompt, and collaborate alongside AI systems. This includes prompt architecture, output interpretation, tool selection, and the "fusion skills" that sit at the interface of human intent and machine capability (Daugherty and Wilson, 2018).

Domain 2: Analytical Reasoning and Critical Thinking. The cognitive discipline required to evaluate AI outputs, identify confabulation and bias, calibrate trust, and challenge machine-generated recommendations. Grounded in dual-process theory and the metacognitive skills that distinguish reliable from unreliable human-AI collaboration (Kahneman, 2011).

Domain 3: Situational Judgement and Exception Handling. The capacity to determine when AI delegation is safe, when human intervention is necessary, and when exceptions must be escalated. Draws on naturalistic decision-making research to build the pattern recognition and response capability that autonomous AI workflows demand (Klein, 1998).

Domain 4: Cross-Domain Generalist Abilities. The breadth of knowledge and cognitive flexibility needed to supervise AI systems that operate across multiple business functions and organisational boundaries. Generalists consistently outperform narrow specialists in contexts that require adaptation and cross-domain synthesis (Epstein, 2019).

Domain 5: Social and Emotional Intelligence. The interpersonal capabilities that become more important, not less, as technical execution shifts to machines. Empathy, trust-building, and psychological safety become the primary differentiators of team performance in AI-augmented organisations (Goleman, 2020).

Domain 6: Strategic Oversight and Ethical Governance. The responsibility to ensure AI systems remain accountable, unbiased, and aligned with organisational values. Encompasses fairness, transparency, privacy, regulatory compliance, and the proactive ethical reasoning that the EU AI Act now makes a legal requirement (Floridi et al., 2018).

Why these domains are interdependent

Each of these domains contributes something the others cannot provide. An orchestrator with strong AI fluency but weak critical thinking will generate outputs efficiently while failing to catch the errors that undermine their quality. Strong critical thinking without situational judgement produces an orchestrator who scrutinises AI outputs rigorously but cannot make rapid decisions about when to intervene in live workflows. Generalist breadth without social intelligence produces cross-functional capability that is technically sound but humanly disconnected. Ethical governance without fluency produces oversight that is principled but operationally uninformed.

The interdependence matters for development planning. Organisations that invest heavily in AI fluency training while neglecting critical thinking, situational judgement, and ethical governance are building a capability profile that is both incomplete and potentially dangerous. The most effective development approaches address all six domains, calibrated to the current level and role of the individuals involved.

New roles, new structures

The emergence of orchestration as a distinct human capability has implications for job design, career pathways, and organisational structure. Roles that were previously defined by the tasks they executed need to be redefined around the orchestration capabilities they require. Performance management frameworks need to assess AI collaboration capability alongside technical expertise. Career progression needs to reward the development of cross-domain breadth and ethical reasoning, not just functional specialisation.

Jarrahi (2018) observes that the most effective human-AI systems are those designed with deliberate attention to the complementarity between human and machine capabilities. This requires organisational leaders who understand both what AI can do well and what humans must continue to do: not as a permanent division of labour but as a dynamic allocation that shifts as AI capability evolves.

How to use this series

Each of the six capability domains is explored in depth in a dedicated article in this series. Each article provides the research foundation for the domain, a practical framework for assessment and development, reflection prompts for practitioners, and recommendations for both individual and organisational development.

We recommend using the [Orchestration Capabilities Assessment](/diagnostic/orchestration-capabilities) alongside the series. The 30-question self-assessment generates a personalised capability profile that identifies your development priorities across all six domains. Use the results to focus your reading on the articles most relevant to your current development needs.

References

Agrawal, A., Gans, J. and Goldfarb, A. (2022) Power and Prediction: The Disruptive Economics of Artificial Intelligence. Boston: Harvard Business Review Press.

Daugherty, P.R. and Wilson, H.J. (2018) Human + Machine: Reimagining Work in the Age of AI. Boston: Harvard Business Review Press.

Dell'Acqua, F. et al. (2023) 'Navigating the Jagged Technological Frontier', Harvard Business School Working Paper, No. 24-013.

Epstein, D. (2019) Range: How Generalists Triumph in a Specialized World. London: Macmillan.

Floridi, L. et al. (2018) 'AI4People: An Ethical Framework for a Good AI Society', Minds and Machines, 28(4), pp. 689-707.

Goleman, D. (2020) Emotional Intelligence: 25th Anniversary Edition. London: Bloomsbury.

Jarrahi, M.H. (2018) 'Artificial Intelligence and the Future of Work', Business Horizons, 61(4), pp. 577-586.

Kahneman, D. (2011) Thinking, Fast and Slow. London: Penguin.

Klein, G. (1998) Sources of Power: How People Make Decisions. Cambridge, MA: MIT Press.

Mollick, E. (2024) Co-Intelligence: Living and Working with AI. New York: Portfolio/Penguin.

Susskind, D. (2020) A World Without Work. London: Allen Lane.

World Economic Forum (2025) Future of Jobs Report 2025. Geneva: World Economic Forum.