The Unlearning Imperative
Why Growth Mindset Is the One Leadership Capability That Makes All Others Possible in the Age of AI
AI is forcing leaders to confront a harder question than adoption: what must they unlearn? This essay argues that growth mindset is no longer a soft leadership trait. It is the meta-capability that determines whether leaders, teams, and boards can keep updating their judgment fast enough to matter.
The Premise: Why Growth Mindset Is Not a Soft Skill
Growth mindset has a branding problem. Somewhere between Carol Dweck’s original research and its adoption by corporate learning and development departments, the concept got flattened into motivational wallpaper. It became something you hang on a conference room wall or list as a hiring criterion, not something you measure, build, or hold leadership accountable for. That dilution has real costs, because what Dweck actually described is not an attitude. It is, as I would frame it, a cognitive operating system that determines how an individual processes failure, uncertainty, and new information; Dweck herself describes it as a belief system about whether abilities can be developed [1], though the framing of mindset as a “cognitive operating system” is the author’s own characterization, not Dweck’s terminology.
In a stable operating environment, the distinction between having a growth mindset and performing one is manageable. Leaders could afford to carry outdated mental models about their industry, their competitors, or their own capabilities, because the pace of change allowed for gradual updating. That condition no longer holds. The emergence of AI as a general-purpose technology has compressed the half-life of domain expertise and executive judgment alike. What worked as a decision framework a few years ago may be actively misleading today.
This is what I mean by the unlearning imperative. The primary leadership challenge is no longer acquiring new knowledge. Most senior leaders have access to more information than they can process. The challenge is identifying which existing beliefs, models, and reflexes need to be retired, and having the cognitive and psychological architecture to actually retire them. Unlearning is harder than learning. It requires confronting the possibility that past success was built on premises that no longer apply, which is an uncomfortable proposition for anyone who earned credibility through demonstrated competence.
The research is directional on this point. A January 2026 Deloitte report found that human capabilities such as curiosity and resilience are differentiating factors in high-performing teams, including teams operating in AI-augmented environments [2]. Separately, McKinsey research suggests that digital fluency alone may be insufficient for effective leadership in the AI age, and that what matters is the capacity to integrate new information into judgment, not just to access it [3].
Learning agility, the measurable behavioral expression of growth mindset, is increasingly recognized as a first-order leadership capability rather than a developmental nice-to-have. That framing is correct, and it is overdue.
The argument I want to make in this paper is more specific: growth mindset is not one leadership capability among many. It is the meta-capability that determines whether every other capability compounds or stagnates. In an AI-accelerated environment, that distinction is no longer academic.
What AI Actually Changes About Leadership (And What It Does Not)
The distinction worth making at the outset is between what AI does to the work of leadership and what it does to the requirements of leadership. These are not the same thing, and conflating them produces both bad strategy and bad governance.
On the work side, the changes are real and consequential. AI compresses execution timelines, automates significant portions of analytical and administrative labor, and accelerates the speed at which decisions need to be made and communicated. Task automation is not coming. It is already here, and it is restructuring the division of cognitive labor inside organizations; a shift reflected in McKinsey Global Institute’s 2023 estimate that current generative AI and other technologies have the potential to automate work activities that absorb 60–70% of employees’ time [4]. Leaders who treat this as a future-state planning problem are already behind.
But here is where the hype diverges from the evidence. Task automation is not judgment replacement. The leadership functions that actually determine organizational outcomes under conditions of ambiguity, conflict, and change are not the functions AI replicates. Contextual moral reasoning, the ability to hold competing stakeholder interests in productive tension, navigating an organization through uncertainty without losing institutional trust: none of these are algorithmic. They require something that, I would argue, AI is not designed to provide: genuine accountability to consequences. A model optimizes outputs. A leader owns them.
Deloitte’s research on high-performing AI-powered teams identifies curiosity, resilience, informed agility, emotional and social intelligence, divergent thinking, and connected teaming as differentiating human capabilities [2], not technical fluency alone, and not prompt engineering. The human skills that compound team performance in AI-augmented environments are the same skills that have always separated good leaders from capable managers. AI has not changed that calculus. It has sharpened it.
There is a governance dimension here that deserves more attention than it typically receives. NACD’s Survey Analysis: Technology Oversight, from the 2024 Public Company Board Practices and Oversight Survey (Aug. 28, 2024), suggests boards are discussing technology more frequently and that frequent discussion correlates with higher confidence in understanding transformative technologies [6]. That pattern is not just a data point. It is a mental model issue. Boards that have not updated their understanding of how AI restructures competitive dynamics are not positioned to ask the right questions about leadership development investment, and the organizations they govern will reflect that blind spot.
The core challenge, then, is not that AI makes leadership easier or obsolete. It is that AI compresses timelines and amplifies execution velocity in ways that make fixed mental models dangerous. Leaders who stopped updating their assumptions about how organizations learn, adapt, and compete will not fail slowly. They will fail fast, and the failure will look like an execution problem when it is actually a cognitive one.
The Anatomy of Unlearning: What Growth Mindset Actually Requires
Growth mindset, properly understood, is not a disposition toward optimism. It is an operational capacity that, for the purposes of this paper, I break into three distinct components: resilience under pressure, adaptability amid structural uncertainty, and learning agility in genuine ambiguity. Each of these is measurable, developable, and consequential. Treating them as a unified abstraction is one reason the concept gets reduced to motivational language instead of management architecture.
Resilience here means the capacity to maintain decision quality and judgment coherence when conditions are adverse. Not cheerfulness. Not tolerance for discomfort. Actual performance stability under pressure, which is a trained capability, not a temperament.
Adaptability is the structural dimension. It describes a leader’s ability to revise operating assumptions when the environment changes in ways that invalidate prior frameworks. LinkedIn’s 2024 Work Change Snapshot found that executives report an accelerating pace of workplace change and widespread adaptation of leadership styles [7], which is often cited as evidence of instability. The more accurate framing is that continuous role redefinition is now the standard operating condition. Adaptability is what allows leaders to function in that condition rather than resist it.
Learning agility is the most precisely defined of the three, and the most misunderstood. It is not general intelligence. It is not domain expertise. It is not curiosity as a personality trait. Learning agility is the capacity to extract transferable insight from novel experience and apply that insight in new situations [5]. I would add a further criterion: genuine agility requires doing so under time pressure, before the environment provides confirmation that the application was correct. That last clause matters. Applying a lesson after the results are in is not agility. Applying it in real time, without the safety of precedent, is.
What makes this hard is not acquisition. Leaders at the senior level are generally capable of absorbing new information. The difficulty is unlearning, and this is where the psychology becomes material. Discarding a prior mental model is cognitively harder than forming a new one, because the prior model is not inert. It is load-bearing. It is connected to prior decisions, public commitments, and in many cases, professional identity.
For credential-based leaders, the cost of unlearning is particularly high. When authority derives substantially from demonstrated expertise in a specific domain, revising that domain’s governing assumptions carries an identity threat that is not symbolic. It is structural. The leader who built their credibility on knowing how things work faces the highest psychological friction when the mechanism changes. This is precisely why learning agility is scarce at senior levels, and why its absence is so consequential when conditions shift.
Growth Mindset as Organizational Architecture, Not Personal Virtue
Growth mindset fails organizations not because the concept is wrong, but because it gets assigned to the wrong unit of analysis. When it lives only in the leader’s psychology, it is invisible to the organization, unmeasurable in practice, and incapable of surviving personnel transitions. The shift required is from personal virtue to systemic infrastructure, and that shift has specific operational components.
The high-performer execution model offers the clearest entry point. In practice, high-performing leaders often establish metrics reviews early, define decision rights explicitly, and create accountability structures before those structures are needed under pressure (practitioner observation). These are not administrative preferences. They are the structural preconditions that allow organizational learning to occur at speed rather than through crisis. Without them, learning happens accidentally, and adaptation is reactive rather than designed.
Kotter’s 8 Accelerators, introduced in Accelerate (2014), provide a practical framework for making this concrete. The Accelerators were developed precisely to address the problem of treating change as episodic rather than continuous. The dual operating system they describe, in which a network of volunteers runs alongside the formal hierarchy, is a structural mechanism for institutionalizing adaptive capacity [8]. It is not a culture initiative. It is an architecture decision. Applied to AI adoption, the implication is straightforward: organizations need ongoing structural support, not episodic change programs.
The 12-to-36-month window following a leadership transition or major organizational inflection point is where this choice becomes consequential. During that window, leaders either build the infrastructure for organizational learning or default to activity. Activity is measurable, visible, and politically comfortable. Infrastructure takes longer and is harder to defend in quarterly reviews. But the compounding effect is asymmetric: organizations that build learning infrastructure in that window absorb subsequent disruptions more effectively. Those that bypass it spend years managing the downstream consequences of having optimized for execution without building the capacity to revise the assumptions underneath it.
The AI governance context makes this concrete. While leading organizations navigating complex technology commitments, I’ve observed repeatedly that the leadership behavior problem precedes the technology problem. Organizations that lack learning infrastructure cannot operationalize AI commitments regardless of tool investment. Deloitte’s research confirms that human capabilities including curiosity, resilience, and emotional intelligence become especially valuable in AI-augmented teams [2]. Purchasing technology without building the leadership architecture to absorb it produces activity, not capability.
The Communication Dimension: Authenticity as a Growth Mindset Signal
The communication practices of a leader are not separate from their leadership philosophy. They are the most visible expression of it. Growth mindset, to the extent it exists in a leader at all, shows up first and most legibly in how that leader communicates under conditions of uncertainty.
The specific behavior that matters is public narration of revised thinking. When a leader says, in front of their team or their board, “here is what I believed, here is what changed my assessment, and here is where I now stand,” they are doing something structurally significant. They are demonstrating that updating a position is a sign of analytical rigor, not a concession of weakness. That demonstration, repeated consistently, becomes the cultural signal that permission exists to learn out loud. Without it, the organization takes its cue from the leader’s apparent certainty and begins suppressing the candid input that learning requires.
This is where AI introduces a specific and underappreciated risk. AI can legitimately help a leader organize ideas, improve clarity, and stress-test structure. What it cannot do is generate the authentic viewpoint that makes communication credible [3]. When leaders allow AI to write their core messages rather than pressure-test their own thinking, they eliminate the one element that builds organizational trust over time: the evidence that a real person worked through the problem and arrived at a considered position. The polished output may be indistinguishable from genuine reflection. The relationship it builds is not.
This matters most in contexts where the subject matter itself is still being understood. AI governance is the clearest current example. Formal governance frameworks are still evolving alongside the technology itself: the EU AI Act entered into force in 2024, NIST released the AI RMF 1.0 in 2023, and ISO 42001 followed that same year [9]. A leader who communicates false confidence in that environment does not appear authoritative. They appear uninformed. Transparency about what a leader does not yet know is not a credibility liability. It is the only credible posture available.
The same principle applies to mentorship. When leaders narrate the process of reasoning through ambiguity, not just the conclusion they reached, they are performing a developmental act with organizational-scale consequences. The team learns not just what the leader decided but how the leader thinks. That transmission multiplies growth mindset capacity across the organization in a way that no training program or stated value replicates. It is the communication equivalent of showing your work, and in a period of genuine uncertainty, it is one of the most consequential things a leader can do.
Why Growth Mindset Makes Every Other Leadership Capability Compound
The most underappreciated property of growth mindset is not what it does on its own. It is what it does to everything else.
Strategic thinking, stakeholder management, execution discipline, and change leadership are each valuable in isolation. But each of them operates on a model, a set of assumptions about what is true, what matters, and what works. When those assumptions are correct, the capability performs as expected. When the assumptions are outdated or wrong, the capability runs confidently in the wrong direction. Growth mindset is the mechanism by which the underlying model gets updated. Without it, every other leadership capability is only as good as its last calibration.
Consider strategic planning in the current environment. McKinsey’s research on leadership in the age of AI identifies judgment as one of the capabilities that cannot be automated [3], but judgment is not a static possession. It is a function of how frequently and rigorously a leader tests their mental model against new information. A leader who cannot revise their criteria as the competitive landscape shifts is not exercising judgment. They are executing a prior assessment. In a market where AI is compressing team output, resetting cost structures, and altering the economics of competitive advantage, the leaders who will sustain strategic relevance are those who can reassess assumptions in real time, not those who execute historical strategy with confidence.
The mentorship dimension compounds this further. One executive who models adaptive learning, and who actively develops it in three direct reports, does not create three more capable people. They seed nine additional nodes of institutional judgment across the organization, in meetings, decisions, and hiring choices where the executive is not present. Deloitte’s research on high-performing AI-era teams found that curiosity and informed agility are among the capabilities that most distinguish effective teams in technology-augmented environments [2]. Those qualities do not emerge from policy. They propagate from leadership behavior.
The failure mode is visible in organizations that use growth mindset vocabulary while exhibiting fixed mindset behavior. When a leader calls every initiative a learning experience but consistently protects relationships over performance, or conflates activity with progress, the vocabulary becomes a liability. It signals that language and behavior are decoupled, which is one of the more corrosive signals an organization can receive.
The competitive asymmetry is durable precisely because it is behavioral. Competitors can replicate tool adoption. They cannot easily replicate a leadership culture that has built the habit of updating faster than the environment changes.
Practical Architecture: Building Growth Mindset Into Leadership Practice
Growth mindset without structure is aspiration. The leaders who actually embed adaptive learning into organizational life do not rely on personal resolve. They build it into cadences, decision rights, and relationship architecture that function regardless of mood or market conditions.
The three highest-leverage interventions below that I have seen at the senior level are each relational and rhythmic, not episodic.
Deliberate after-action review of revised assumptions. This is distinct from standard project retrospectives or “post-mortems”. The specific question is: what did we believe going in that we no longer believe coming out, and what caused the update? Run this quarterly at the executive team level. Document it. Over time, the pattern of assumption revision becomes a diagnostic instrument. Teams that cannot identify revised assumptions are not learning. Teams that can identify them but only in hindsight are lagging. Teams that revise assumptions in real time, mid-execution, are operating with genuine learning agility.
Public modeling of learning in executive communications. When a senior leader states, in writing or in a town hall, that they changed their mind and explains why, it produces a disproportionate cultural signal. Most leaders treat changed positions as vulnerabilities. The evidence runs the opposite direction. Curiosity and resilience are reinforced when leaders visibly model learning behavior [2]. The practical implication: build a standing agenda item into quarterly all-hands or leadership communications that surfaces one significant belief revision from the executive team.
Mentorship structured to reward challenge. Traditional mentorship flows expertise downward. The more durable learning design inverts part of that relationship, creating explicit permission and expectation for the mentee to surface assumptions the leader has not examined. This is not a soft arrangement. It requires the leader to formalize what questions they want pushed back on, and to create accountability for whether those challenges actually occurred. The most durable leadership development gains come from structured relationships and stretch assignments, not from formal training programs in isolation [10]. Allocation of development resources should reflect that reality.
At the board level, growth mindset stress-testing belongs inside risk review cycles, consistent with the broader technology-oversight focus reflected in NACD survey work [6]. Scenario planning sessions that incorporate AI disruption and regulatory shift scenarios should include an explicit assumption audit: what beliefs underpin each scenario, and which of those beliefs have not been tested in the past twelve months. That framing converts scenario planning from narrative exercise into a genuine adaptive learning mechanism.
The Unlearning Imperative: A Closing Argument
Every intervention described in this paper, from deliberate assumption audits to the rhythmic after-action reviews that close the previous section, rests on a single premise. Growth mindset is not a supplement to leadership capability. It is the precondition that determines whether capability investments return value or depreciate over time.
The evidence accumulated across these sections points in one direction. McKinsey’s work on leadership in the AI era confirms that human judgment, strategy, and creativity remain irreplaceable precisely because they are adaptive, not because they are static [3]. Deloitte’s findings on high-performing teams identify curiosity and resilience, not technical credentials, as the differentiating factors [2]. Kotter’s accelerator model for change leadership is built around sustained urgency and broad coalition [8], neither of which is possible inside an organization whose leadership models certainty over inquiry. These are not soft findings. They are structural observations about what separates organizations that learn from organizations that calcify.
The direct challenge to senior leaders is this: the question is not whether you believe in continuous learning. Most leaders say they do. The operative question is whether your operating cadences, your communication behaviors, and your development investments make that belief structurally visible to the people you lead. Belief without architecture is intention without consequence. The people you lead are watching the delta between what you say about learning and how you respond when you are wrong.
Legacy, in this context, is not about accumulated expertise. The leaders whose judgment outlasts their tenure are those who built institutional learning infrastructure, not those who arrived at certainty and held it. They left behind organizations capable of revising assumptions without crisis, absorbing new information without fragmentation, and developing the next layer of leadership without replicating fixed mental models.
The unlearning imperative is not philosophical. In any environment where the cost of a fixed mental model can be measured in missed opportunity, eroded trust, or strategic misstep, the capacity to revise, to discard, and to rebuild is an operational requirement. That cost is not hypothetical. It is visible in organizations that over-committed to AI without governance infrastructure, in leadership teams that defended strategies past their useful life, and in cultures where candor was structurally impossible.
Build the architecture. Run the cadences. Signal the behavior. The organizations that do will compound. The ones that do not will mistake stability for strength until the gap becomes undeniable.
References
[1] Carol S. Dweck, Mindset: The New Psychology of Success, Random House, 2006.
[2] Deloitte, “Human Capabilities Are at the Heart of High-Performing Teams,” Deloitte Insights, Jan. 14, 2026.
[3] McKinsey & Company, “Building Leaders in the Age of AI,” McKinsey Quarterly, Jan. 12, 2026.
[4] McKinsey Global Institute, “The Economic Potential of Generative AI: The Next Productivity Frontier,” June 14, 2023.
[5] Center for Creative Leadership, “Tips for Improving Your Learning Agility,” Sept. 2, 2025.
[6] National Association of Corporate Directors, “Survey Analysis: Technology Oversight,” 2024 Public Company Board Practices and Oversight Survey, Aug. 28, 2024.
[7] LinkedIn Economic Graph, “Work Change Snapshot,” 2024.
[8] John P. Kotter, Accelerate: Building Strategic Agility for a Faster-Moving World, Harvard Business Review Press, 2014.
[9] European Commission, “AI Act Enters into Force,” Aug. 1, 2024; NIST, “AI Risk Management Framework 1.0,” Jan. 26, 2023; ISO/IEC 42001:2023, Artificial Intelligence Management Systems.
[10] Center for Creative Leadership, “The 70-20-10 Rule for Leadership Development,” Apr. 24, 2025.