The Risks of Cognitive Delegation and the Challenge of Accountability in the Age of AI

• AI is reshaping how investors develop and exercise professional judgment.
• Greater reliance on AI can erode critical thinking and blur accountability for investment decisions.
• Competitive advantage will increasingly depend on preserving human judgment and accountability as the guiding force in the synthesis of human and machine intelligence.

Warren Buffett once said, “The most important quality for an investor is temperament, not intellect.” This insight carries significant implications in the age of artificial intelligence (AI): in investing, success depends on the ability to exercise sound judgment under conditions of uncertainty. Yet the adoption of AI is transforming the profession in ways that may violate this principle.

The integration of AI into the investment process requires a reconfiguration of how knowledge is acquired, validated, and ultimately translated into investment decisions. While considerable attention has already been devoted to the efficiency gains that AI offers in acquiring and synthesizing data, far less scrutiny has been applied to the impact of AI on the formation and preservation of human judgment, a foundational capability on which the investment profession depends.

As AI becomes deeply embedded in investors’ daily workflows, it is reshaping not only how investment analysis is executed, but also how professional judgment is developed, sustained, and applied, raising questions about the cognitive capabilities investors may lose and who remains accountable for the decisions that result.

This post, part of a quarterly series examining AI’s impact on investment management, draws on frontline experience and insights from investment specialists, academics, and regulators collaborating on the bi-monthly newsletter The Augmented Intelligence Investor.¹ We examine two critical lessons: the risks AI poses to investors’ cognitive judgment and the importance of accountability as AI plays an increasingly central role in investment processes.

The investment industry faces a situation whereby the same systems that enable analytical efficiency also facilitate the outsourcing of cognition. In practical terms, this leads to a growing tendency among investment professionals to rely on AI-generated outputs prior to developing their own sufficiently robust internal understanding of the underlying analytical processes. Recent research clearly shows that such behavior can introduce significant fragility into the investment process (Gerlich, 2025; Jose et al, 2025; Lenhardo, 2026; Strömberg et al, 2026). While investment theses may appear technically coherent, investors risk losing the ability to properly question, defend, and adapt them when necessary.

Investment management has always required sound judgment in the face of uncertainty and incomplete information. Historically, this judgment has been developed through experiences with complex environments and the evaluation of evidence under stressful conditions. Although inherently inefficient, these processes remain the primary means by which investors build tacit knowledge and expertise. In contrast, current AI systems are specifically designed to eliminate such friction. In doing so, they compress, and potentially bypass, the pathway through which investment expertise has traditionally been acquired.

This may have significant implications for talent development within investment organizations. Entry-level professionals, who have traditionally built their expertise through a series of increasingly demanding analytical tasks, can now produce sophisticated outputs, such as financial models, investment theses, and risk assessments, without fully internalizing the underlying conceptual frameworks. Over time, this may create a generation of analysts whose investment theses mask gaps in their foundational understanding. This imbalance becomes particularly evident in live discussions and decision-making settings, where the ability to defend assumptions, address counterarguments, and revise conclusions in real time remains critical.

Importantly, this phenomenon is not limited to junior professionals. Once cognitive delegation becomes normalized, it also affects experienced investment professionals. As reliance on AI-assisted outputs grows, the maintenance of internal mental models (i.e., the simplified yet essential frameworks investors use to interpret complex realities) may gradually erode. This erosion introduces a significant risk into the investment process, particularly when investors face high-stakes situations where time pressure often limits thorough verification and where independent reasoning is most critical.

Another important point is the critical role of accountability. As investment processes become increasingly integrated with AI, the line between human judgment and machine inference continues to blur. Yet history has shown that society at large, together with legal, regulatory, and ethical frameworks, holds individuals and organizations, not machines, responsible for outcomes – despite the obvious potential of AI to generate a competitive advantage due it its capability to collect and process much larger quantities of information than a single human brain. This reality presents investment firms with a pressing governance challenge. They must align their control frameworks by treating AI not as an autonomous decision-maker, but as a tool that operates within a well-defined architecture of human oversight. Such a framework requires clear decision rights, transparent documentation of analytical processes, and robust audit mechanisms capable of tracing how outputs are generated and incorporated into final decisions. Without these safeguards, investment firms risk falling into black-box dependency, where conviction rests on outputs that are neither fully understood nor properly challenged. This exposes investment professionals, investment firms, and their clients to serious financial, reputational, and regulatory consequences.

Compounding this issue is the growing difficulty of performance attribution within investment firms. The investment profession has long relied on the ability to link outcomes to individual contributions in order to align incentives with genuine capability. As AI becomes deeply embedded in the analytical workflows, however, disentangling the respective contributions of humans and AI becomes increasingly complex. This ambiguity complicates not only compensation structures and bonus allocations but also promotion decisions, and the broader talent evaluation processes. Over time, the inability to clearly attribute results may weaken existing incentive schemes, erode trust in traditional performance assessments, and make it harder for investment firms to identify, develop, and reward true investment skills. In an industry where demonstrated judgment under uncertainty has long been the ultimate measure of success, this erosion of attribution clarity represents a significant risk.

As AI expands the volume of available information and accelerates the speed of analysis, competitive advantage for investment firms will increasingly depend not on access to data or computational power, but on judgment and accountability. Firms best positioned to thrive will treat both as critical, scarce resources, designing investment processes that preserve constructive friction, foster independent thinking, and challenge both human and AI-generated conclusions.

They will view AI not as a substitute for human expertise, but as a powerful tool whose effectiveness depends on the quality of the human systems in which it operates. In this sense, a thoughtful adoption of AI elevates the relevance of human cognition in decision designs to a new level. On this new level, investment managers spend less time on informational legwork and more on scrutinizing their own decision heuristics. Ultimately, this should enable them to make an informed decision on where and how they can incorporate AI into their own decision processes.

References

Gerlich, M. (2025), From Offloading to Engagement: an Experimental Study on Structured Prompting and Critical Reasoning with Generative AI. Data 2025, 10, 172. https://doi.org/10.3390/data10110172

Jose B, Joseph D, Mohan V, Alexander E, Varghese SK and Roy A (2025) Outsourcing cognition: the psychological costs of AI-era convenience - PMC 

Lenhardo, M (2026), Is AI ruining our skills? Early results are in — and they’re not good | Nature

Strömberg, D, V Lei and Y Wu (2026), ‘DP21577 The Generative AI Learning Penalty: Evidence from Chinese Secondary Education‘, CEPR Discussion Paper No. 21577. CEPR Press, Paris & London. https://cepr.org/publications/dp21577