_edited.jpg)
Who will decide? Who will make the choice to delegate, or not delegate, responsibility and authority?
February, 15 2026
Pamela Woitschach
The original title for this piece was “Human in the Loop Is Tactical. Human in the Lead Is Strategic.”
That was until I arrived at the last section of the book: Genesis: Artificial intelligence, hope, and the human spirit. From Kissinger, H. A., Mundie, C., & Schmidt, E. (2024).
On page 217, the authors ask: Who will decide? Who will make the choice to delegate, or not delegate, responsibility and authority? The paragraph continues with additional questions that I will address in a separate post on the Intelligent Leadership Framework.
For now, I remain with the more foundational question: who is in the lead?
Julie Sweet, in the last few second of this panel (42:22) invited us to transform the cognitive narrative from a passive to an active approach moving:
From Human in the Loop -------> to -------> Human in the Lead.
In a previous post, I described powerful AI as “a country of geniuses in a datacenter,” borrowing from Dario Amodei (2024) in Machines of Loving Grace.
Henry Kissinger, Craig Mundie, and Eric Schmidt (2024) wrote that "because of its unique methods of inquiry and learning, AI will be capable of inhuman achievements in both scale and resolution, thereby activating changes fundamentally different from any previous human invention."
What Amodei and the authors of Genesis describe is not incremental progress but structural acceleration.
Suleyman, M. (2023). The coming wave: Technology, power, and the twenty-first century’s greatest dilemma, describes "a world in which centuries of technological change compress into years and consequences ricochet globally in seconds."
If millions of autonomous intelligent systems operate continuously across sectors and geographies, at superhuman speed and competence, inserting a human into every meaningful decision point becomes structurally infeasible.
It is not about lowering standards either. Kissinger, Mundie, and Schmidt (2024) caution that humanity should not embrace weaker moral standards even if technological logic tempts us toward them.
From all the above, Human in the loop while operationally relevant, becomes not manageable or scalable.
Human in the Lead
Despite companies exploring the use of AI, processes are still far from being scaled. Only one-third is scaling AI solutions, and only a small number of respondents are fully scaled. Says the Quantum Black AI Report by McKinsey and Company, 2025, recent survey (p. 4-5).
Julie Sweet has argued that human in the lead is the only scalable (governance and oversight) model.
Human in the lead can be analyzed across three levels: macro, meso, and micro
At the macro level: Human in the Lead concerns structural power and normative direction. Humans design and build models and platforms, define the institutional and regulatory environments in which they operate, and embed social, economic, and cultural values into these systems. At this level, the question is not supervision but authorship.
At the meso level: Human in the Lead operates through organizations and institutional structures. Here, Human in the Lead is exercised through strategic coordination, policy design, risk management architectures, and governance mechanisms.
At the micro level: Human in the Lead remain central in developmental and operational design of processes. Even in highly automated pipelines, human actors direct where attention, resources, and applications of AI and prospective AGI are focused. Micro level Human in the Lead concerns design choices and deployment decisions, not merely oversight.
Human in the Loop does not disappear under this piece. It reflects a more operationally ingrained in day-to-day activities. It refers to a human actor embedded within the workflow of an AI system. It remains necessary for auditing, accountability, and exception handling. However, it cannot serve as the primary control mechanism in a world of fast, autonomous, massively parallel intelligence.
If a system runs millions of parallel processes without fatigue or delay, the throughput mismatch renders loop based supervision non scalable by definition, a point repeatedly emphasized by Mundie. Kissinger, Mundie, and Schmidt, warns that AI outcomes may increasingly exceed human interpretability. For me, this distinction is central to why Human in the lead is critical.
Human in the Lead is not only about scalability. It is also about preventing displacement, a concept developed in Genesis, where systems gradually shape human priorities rather than merely executing them.
At a World Economic Forum session, Arjun Prakash stated that AI is not making decisions; it reflects the values society aligns it with. He emphasized that AI amplifies intent and culture. It is not neutral infrastructure but multiplicative infrastructure.
Yet Mundie (2025) also observed that the assumption that a human can write rules to prevent events they cannot anticipate is flawed. The issue is not capability. In many domains, AI already exceeds human performance in speed and precision. The issue is governance under uncertainty.
Consider now some examples of concrete approaches:
Anthropic Claude’s Constitution, embedding explicit normative principles into model training.
Yoshua Bengio (2023) has proposed grounding advanced systems in scientific values such as safety, transparency, and social responsibility, advancing the concept of a Scientist AI that reinforces epistemic integrity rather than undermines it.
At the macro level, Kissinger, Mundie, and Schmidt (2024) propose grounding models including information from all levels of society. States will designate representatives to shape AI’s architectural foundations, sharing and including what is encoded, preserved, and governed. Rather than relying exclusively on prescriptive rulebooks, they explore embedding moral narratives, drawing on the anthropological notion of "doxa", socially transmitted belief structures learned through culture rather than formal codification. They argue that grounding models should reflect international, regional, national, and local laws, as well as community norms.
The authors also cautioned that agreement will not be reached, for what all possible information should be part of grounding models.
Across macro, meso, and micro levels, the argument converges. Humans cannot scale as micro supervisors of superhuman systems. But humans can scale values, institutional architectures, and governance frameworks that channel AI and AGI attention toward the development and facilitation of human missions.
The tension remains that values are not universal. Different societies operate under distinct normative frameworks, but this ambiguity can be a fertile ground in which to cultivate dialogue (Kisssinger, et al. 2024 p.117).
What Kissinger et al (2024) described as world order will shape how AI systems are designed, aligned, and deployed. Genesis closes by suggesting that the traditional equilibrium model of world order may itself be destabilized by AI.
Technologically advanced nations will therefore exert disproportionate influence over emerging architectures of governance.
As Suleyman (2023) notes, given the geostrategic and commercial value of advanced AI, it is implausible that states or corporations will voluntarily relinquish transformative capabilities.
The first step, though not the only one, is to construct interlinked and mutually reinforcing technical, cultural, legal, and political mechanisms capable of maintaining meaningful human authority during exponential technological change.
The central question remains unawsered: but as Kissinger, Mundie, and Schmidt (2024) noted, “Today’s human leaders should prepare to be the first in a line of human sovereigns to face struggle of locating a balance between leveraging the advantages, and in some cases, the need for AI governance without going so far as to succumb total dependency.” p. 107-108.
References
Amodei, D. (2023). Machines of Loving Grace: Artificial General Intelligence and the Future of Human-Machine Collaboration. [Essay]. Retrieved from https://darioamodei.com/essay/machines-of-loving-grace
Bengio, Y. (2023). AI Scientists: Safe and Useful AI?. Retrieved from https://yoshuabengio.org/2023/05/07/ai-scientists-safe-and-useful-ai/
Kissinger, H. A., Mundie, C., & Schmidt, E. (2024). Genesis: Artificial intelligence, hope, and the human spirit. Little, Brown and Company.
McKinsey & Company. (2025). State of AI in 2025: Agents, innovation, and transformation (Quantum Black AI). McKinsey & Company. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
Mundie, C. (2025, January). Genesis: artificial intelligence, hope, and the human spirit | LSE Event. Retrieved from https://www.youtube.com/watch?v=48YvBDaqO-E
Suleyman, M. (2023). The coming wave: Technology, power, and the twenty-first century’s greatest dilemma. Crown.
WEF 2026
Comments