Break Everything and Start Fresh - Perspectives on Leadership in Implementing LLMs
We crave the new, yet cling to the familiar. In the age of AI, leaders must rise above this instinct to drive meaningful organizational change
From an evolutionary standpoint, we are wired for an inherent tension between seeking the new and preserving the status quo. In other words, being either hyped or skeptical about new technologies is possibly as human, as it gets.
Consider our aptitude towards certain cognitive biases like the novelty bias (a preference for new stimuli) and the status quo bias (a preference for maintaining the current state). Consider also the sheer volume of novelty nowadays, which creates conditions very different from our evolutionary past, while we still maintain our evolutionary response. Humans are not naturally equipped to regulate exposure to constant novelty, which can lead to overstimulation, decision fatigue, or even addiction to newness. So, maintaining a balanced approach within the opposing tendencies of hype and scepticism is challenging and context-dependent. And this is exactly what differentiates good leadership in the age of AI.
See, the more we think about implementing LLMs at the leadership level, the more we begin to realize that the biggest challenges there are not technical but deeply human ones. And leadership in the context of hybrid human-machine teams becomes increasingly complex, involving reshaping organizational culture, managing expectations, and balancing innovation with caution. But can you do this with existing organisational structures?
In our latest technical panel discussion in 2024, we talked about the challenges to LLM implementation, but our conversations almost always boiled down to a central question: What does leadership look like today?
To explore these challenges, we’ve gathered an amazing panel of expert practitioners to discuss multidisciplinary perspectives on how organisations and leaders, in particular, can navigate the complexities of LLM deployment. Our panelists were:
Martina Mendola, Research Associate Manager, Accenture The Dock. Dr. Martina Mendola leads research in the Human Sciences Studio at Accenture’s Global Innovation Center, The Dock.
Saskia Lensink, NLP and LLM Specialist, TNO (Netherlands Organisation for Applied Scientific Research). As a consultant and business developer at TNO, Saskia Lensink specialises in language and speech technologies.
Amir Feizpour, Founder & CEO, Aggregate Intellect. Amir Feizpour leads Aggregate Intellect, an AI platform geared toward building a “generative business brain” for science- and service-based companies.
Perhaps one of the boldest insights came from Amir’s assertion that leaders might need to “break everything down to the ground” and rethink organizational structures entirely.
The suggestion to "break everything down to the ground" is, of course, meant as a metaphor for fundamental disruption, where incremental changes might not be sufficient when adapting to an AI-enhanced workflow. At its core, this metaphor highlights that our organizational structures, workflows, and job roles might be inadequate in the context of the collaborative potential of humans and machines.
Being the moderator of this discussion, this got me thinking that LLMs and other AI tools augment human work which necessitates a rethinking of team dynamics, decision-making processes, and accountability structures. And as roles evolve, leaders and employees alike need to develop new skills, including how to effectively collaborate with AI. Leading the adoption of LLMs is one such task that requires more than knowledge and visionary leadership. It requires empathy, kindness humility, curiosity and altruism.
In other words, organisational transformation to LLMs adoption requires visionary leaders who can not only articulate the need for change but also inspire trust and commitment. These leaders must embrace ambiguity, foster collaboration, and prioritize ethical considerations.
Leadership has evolved beyond the outdated notion of viewing employees as mere productivity tools. In today’s complex environment, technical expertise and management skills alone are no longer sufficient. A good leader must also take on the role of a cognitivist, skillfully navigating the delicate balance between hype and scepticism - two sides of the same coin shaped by cognitive biases. This demands a deep understanding of human dynamics, the ability to critically evaluate emerging trends, and the foresight to steer teams through uncertainty with clarity, wisdom, and adaptability.
Martina emphasized the importance of addressing this with behavioural science. “Leaders must understand and respect the mental models people hold about AI - whether they see it as a utopian saviour or a dystopian threat,” she noted. We need to foster a culture of trust. And speaking about culture, leaders must not forget the multicultural aspect of human behaviour.
As Saskia highlighted, LLMs often lack the cultural and contextual nuances needed for effective global use. While fine-tuning can bridge some gaps, foundational issues remain, especially in underrepresented languages and regions. Participatory AI - bringing diverse voices into the development process - offers a promising path forward but requires a significant commitment from organizations, which also means balancing short-term productivity gains with long-term strategic thinking.
Martina stressed the importance of creating “safe spaces for experimentation,” such as pilot projects and sandboxes, where employees can learn to use LLMs without fear of failure. Moreover, leaders need to align workforce goals with technological capabilities to foster trust and reduce the perception gap between management and employees as overhauling organizational systems can disrupt operations and alienate employees if not managed carefully.
So, the question is, what can you do?
Say Goodbye to Micromanagement
The rise of AI, particularly tools like LLMs, demands a shift in leadership styles from micromanagement to empowerment. Leaders who previously relied on direct oversight and granular control must now focus on enabling autonomy within their teams.
This means a transition from being a taskmaster to becoming a facilitator who creates an environment where team members feel empowered to work and have a safe space to learn, experiment, fail, and innovate without fear of judgment.
Say Hello to Inclusivity
Inclusive leadership begins with creating diverse, multidisciplinary teams that encompass a wide range of perspectives, experiences, and cultural backgrounds. AI, by its very nature, thrives on multidisciplinary collaboration, given the broad spectrum of contexts in which it is applied—even within a single organization. These diverse teams play a crucial role in identifying blind spots in AI systems, uncovering biases, and developing solutions that are more equitable and representative of global needs.
By prioritizing inclusivity, leaders ensure that AI becomes a force for good, driving fairness and reducing systemic inequities. This approach is particularly vital when deploying LLMs in underrepresented languages or culturally nuanced contexts, as Saskia highlighted during the panel discussion. Inclusive practices in AI not only enhance the technology's utility but also align it with the values of justice and diversity, ensuring it serves humanity at large.
Learning to Adapt
In this new era, leadership is no longer about dictating directives but about inspiring and guiding teams through transformation.
Adaptive leaders excel at navigating ambiguity and are attuned to the evolving dynamics of hybrid human-AI teams. They recognize that leadership requires a human-centred approach - balancing technical innovation with emotional intelligence, inclusivity, and ethical foresight.
Adaptive leaders foster environments where employees feel secure to explore the potential of AI, learn collaboratively, and address challenges without fear of failure. This approach not only drives innovation but also builds trust, ensuring that employees feel invested in and aligned with the organization’s vision.
Creating Safety to Bring Ballance
In an era defined by rapid technological advancements and the complexities of hybrid human-AI collaboration, the ability to maintain balance has emerged as a critical leadership skill.
As noted earlier, humans are evolutionarily predisposed to oscillate between novelty bias and status quo bias—between the excitement of embracing the new and the comfort of clinging to the familiar. Leaders, however, must rise above these natural tendencies to navigate the fine line between hype and scepticism with clarity and purpose.
Maintaining balance involves critically evaluating the promises of new technologies like LLMs while remaining grounded in ethical considerations and practical realities. It requires fostering innovation without succumbing to blind optimism and addressing legitimate concerns. This balanced approach is essential for creating trust among employees, stakeholders, and society at large.
While humans may not be evolutionarily built for balance in this regard, the capacity for self-reflection, learning, and intentional action provides a way to cultivate such balance. The successful implementation of LLMs requires a blend of innovation, experimentation, and humility. Leaders must embrace the complexity of change while fostering adaptability and resilience within their teams.
As Amir aptly put it, “The ability to learn how to learn will define success in this rapidly evolving landscape.”