A related ethical question concerns ownership: who does intelligence, and the value it creates, belong to? AI systems are trained on vast amounts of human-generated knowledge, yet the ownership of resulting intellectual property remains contested. Several technology firms have begun formalizing contractual arrangements to address this issue, including Anthropic AI’s proposed $1.5B copyright settlement in September 2025, the largest copyright recovery to date⁷. The debate, however, remains unresolved.

Fairness concerns of this nature are not new. Previous industrial revolutions often required workers to train replacements or offshore counterparts, contributing to downward pressure on wages and rising inequality. The distinction today lies in speed and scale. The “worker” being trained— an AI system—is faster, more focused, and operates continuously. The resulting margin expansion disproportionately benefits shareholders, senior executives, and, in some cases, private investors.

While AI governance has been widely discussed, less attention has been paid to how corporate approaches to AI shape organizational culture, talent development, and human capital formation. If AI displaces entry-level and repetitive roles, hiring freezes and reduced early-career opportunities may disproportionately affect new graduates.

This raises questions about how future leaders will be developed without traditional cycles of mentoring, experiential learning, and progressive responsibility. Leadership styles, collaborative learning, workplace norms, and professional networks are all shaped through sustained human interaction. Removing early-career roles risks eliminating a critical rung on the career ladder, with second-order effects on creativity, engagement, and societal development.

Over the past two decades, rapid technological advancement has repeatedly introduced complex social and governance challenges, from data privacy and cybersecurity to ethical accountability and intellectual property theft. Historically, corporate and policy responses have tended to lag innovation, often emerging only after harm has occurred.

We argue that this cycle must be broken. In the case of AI, corporate and policy frameworks should precede, rather than follow, large-scale deployment to mitigate the most severe risks to capital markets and social stability. A pro-worker AI framework may offer a pathway to preserving resilient income and consumption patterns, which remain foundational to long-term market sustainability.

In our early engagements on AI, discussions were broad and fragmented, spanning adoption, governance, education, and controls. To sharpen our stewardship approach, we have focused board-level engagement on three core areas:

1. Whether boards possess the appropriate skills and knowledge to oversee AI-related risks, and whether effective oversight structures are in place
2. How AI adoption has reshaped organizational structure and culture
3. Whether AI-driven productivity gains can be reliably quantified

The quality, composition and effectiveness of boards remain one of the core pillars in which we evaluate the corporate governance practices of portfolio companies. Naturally, evaluating whether boards possess relevant AI expertise—through training or experience—is a foundational step in understanding oversight quality.

We seek to understand how boards intend to evolve, including changes through appointments or retirements, and what skills are being prioritized. With the right skills in place, well-structured, largely independent and high-performing boards can help support the oversight of material risks, including financially material environmental and social risks.

The second line of inquiry related how AI has shaped investee companies gave varied responses in our experience. Some focused on discrete tools and efficiency gains, while others speak at a cultural level. Understanding how AI reshapes corporate culture has proven particularly insightful.

In some of our engagements with portfolio companies, we have seen a higher degree of emphasis on people-centric philosophies but supported or empowered by technology – and more specifically, AI. In some cases, such axioms are embedded into corporate descriptions and communications. These statements represent what MIT Sloan Professor Edgar H. Schein describes as cultural “artifacts”—visible expressions of organizational values⁸. However, looking at artifacts alone are insufficient, and management’s espoused values may also not align with employees’ underlying assumptions. To assess culture meaningfully, stewardship requires engagement across organizational levels, boards, management, and non-management employees, to compare perspectives and identify inconsistencies.

Schein’s model also emphasizes that culture is transmitted through shared learning and the passing on of solutions to new members. That is, this passing on of solutions is a testament to whether such shared solutions or values are valid. In an agentic AI environment, where AI increasingly substitutes for entry level positions, opportunities for this transmission diminish. The implications are twofold. 

First: Accountability remains with human decisionmakers, even as AI performs more analytical work

Second: The developmental feedback loop, where poor work becomes a teachable moment, is weakened.

While re-querying AI systems may be efficient, it removes the process of developing human capital, diminishing long-term organizational capability. More importantly, without structured opportunities to test, reinforce, and evolve shared assumptions, with human participants that can receive feedback and be held accountable, organizational culture may be at risk of erosion over time. 

Although shareholder primacy has long dominated corporate strategy, market outcomes are shaped by broader societal assumptions. As discussed in Part I, the “Polanyi Moment” highlights periods when market systems are forced to re-embed within social structures. While thinkers such as Adam Smith and Milton Friedman shaped prevailing views on profit maximization, these ideas have been challenged.

Stakeholder theory, advanced by Edward Freeman, argues that long-term value creation requires balancing the interests of employees, communities, and the environment alongside those of shareholders⁹. This perspective has gained traction in parts of the developed world, grounded in the belief that unchecked profit maximization may ultimately undermine shareholder value itself— particularly when shareholders are also workers through pension and retirement systems.

As markets continue to reward short-term AI-driven gains, the question arises whether companies are prioritizing rapid wins at the expense of labour, potentially increasing reliance on government transfers or universal income mechanisms.

For investors, understanding AI-driven productivity gains first requires reliable, standardized disclosure. Standardized disclosures surrounding crystallized productivity gains would be helpful for investors to separate efficiencies already had, against hopeful anecdotes of potential use cases which are less certain. Without quantification, it is difficult to distinguish realized efficiencies from aspirational narratives.

Where gains are realized, shareholders are likely to capture a significant share through margin expansion and equity appreciation. Many management executives also participate through equity-based compensation plans, which are often explicitly tied to performance thresholds. In more extreme examples, some manager-owners have even asked shareholders to vote and approve significant share ownership grants tied to extraordinary performance hurdles, locking in at the very least the earning potential tied to productivity gains from AI if astronomical performance-hurdles are met.

This raises the question of whether employees— whose labour supports both production and consumption—should also participate in AI-driven value creation. Employees may face structural disadvantages: continued participation in equity plans is often contingent on employment, while many employees are excluded entirely based on seniority or eligibility.

One potential solution is to reimagine equity participation through ‘employee benefit trust’-type structures, allocating a portion of AI-driven productivity gains to current and former employees, irrespective of ongoing employment. Such structures would be distinct from traditional compensation plans and could be managed similarly to pension arrangements.

While not presented as a definitive solution, this approach illustrates how private-sector innovation could mitigate labour displacement risks more effectively than eventual government intervention.

Governance of such frameworks could also benefit from meaningful employee representation at the board level, balancing shareholder and management interests. International examples, such as employee representation at the board level in German markets¹⁰, demonstrate the viability of such governing structures. More thinking would have to go into the actual level representation of ‘employee representative’ directors, but the idea would be to fairly represent the employee-class at the board, balancing the interests of shareholders and management. Having such representation at the board level would add appropriate perspectives to discussions and company strategy, which helps enhance the ability of the board to think long-term and manage risks.