Three structural themes appear many times in academic papers, business studies and regulatory reports. Taken together, they suggest that AI won’t only improve investor skills. Instead, expertise is reassessed, the importance of process design is increased, and competitive advantage is shifted to those that understand the technical, institutional, and cognitive limitations of AI.
This post is the fourth a part of a quarterly series on AI developments relevant to investment management professionals. It draws on insights from contributors to the bi-monthly newsletter and builds on previous articles to offer a more nuanced have a look at the evolving role of AI within the industry.
Performance exceeds reliability
However, a variety of studies warn that benchmark success masks fragility in real-world scenarios. OpenAI and Georgia Tech (2025) show that hallucinations reflect a structural trade-off: efforts to cut back incorrect or fabricated answers inherently limit a model’s ability to reply rare, ambiguous, or underspecified questions. Related work on causal extraction from large language models also suggests that strong performance in symbolic or linguistic reasoning doesn’t result in robust causal understanding of real-world systems (Adobe Research & UMass Amherst, 2025).
This distinction is crucial for the investment industry. Investment evaluation, portfolio construction and risk management will not be based on stable ground truths. The results are regime dependent, probabilistic and highly sensitive to tail risks. In such environments, results that appear coherent and authoritative but are nonetheless incorrect can have disproportionate consequences.
For investment professionals, because of this AI risk is increasingly much like model risk. Just as backtests routinely overstate actual performance, AI benchmarks are likely to overstate decision reliability. Companies that deploy AI without proper validation, foundation and control frameworks risk embedding latent weaknesses directly into their investment processes.
From individual competence to institutional decision-making quality
The second theme is that AI commercializes investment knowledge while increasing the worth of the investment decision-making process. Evidence from AI use in production environments makes this clear. The first large-scale study of AI agents in production finds that successful deployments are easy, narrowly defined and repeatedly monitored. In other words, AI agents today are neither autonomous nor causally “intelligent” (UC Berkeley, Stanford, IBM Research, 2025). In regulated workflows, smaller models are sometimes preferred because they’re more verifiable, predictable, and stable.
Behavioral research supports this conclusion. The Kellogg School of Management (2025) shows that professionals underuse AI when its use is visible to superiors, even when it improves accuracy. Gerlich (2025) notes that frequent use of AI can reduce critical pondering through cognitive offloading. Therefore, if AI is left unmanaged, there may be a dual risk of underutilization and over-dependence.
For investment organizations, the lesson is due to this fact structural: the advantages of AI don’t accrue to individuals, but somewhat to the investment processes. Leading firms are already embedding AI directly into standardized research templates, monitoring dashboards and risk workflows. Governance, validation and documentation are increasingly more vital than mere analytical firepower, especially as supervisors themselves undertake AI-powered oversight (State of SupTech Report, 2025).
In this environment, the normal idea of the “star analyst” is starting to falter. Repeatability, verifiability and institutional learning can develop into the true source of sustainable investment success. Such an environment requires a major change within the design of investment processes. Following the worldwide financial crisis (GFC), investment processes became largely standardized, with a powerful give attention to compliance.
However, the evolving environment requires optimization of investment processes by way of decision quality. This change is important in scale and difficult to realize since it is determined by managing individual behavioral change as a fundamental level of organizational adaptability. This is something the investment industry has often tried to avoid through impersonal standardization and automation – and now it’s trying again through AI integration, misrepresenting a behavioral challenge as a technological challenge.
Why AI’s limitations determine who unlocks value
The third theme focuses on the constraints of AI somewhat than viewing it as only a technological race. On the physical side, infrastructure boundaries develop into binding. Research shows that only a small portion of announced U.S. data center capability is definitely under construction, with network access, power generation, and transmission schedules measured in years somewhat than quarters (JPMorgan, 2025).
Economic models support why this is vital. Restrepo (2025) shows that in an economy driven by artificial general intelligence (AGI), production occurs linearly by way of computing power, not labor. Therefore, owners of chips, data centers and energy generate economic returns. The placement of computing infrastructure, chips, data centers, energy and platforms that manage allocation is the critical think about value creation by removing labor from the expansion equation.
Institutional constraints also require greater attention. Regulatory authorities are rapidly expanding their AI capabilities, thereby increasing expectations for explainability, traceability and control when using AI within the investment industry (State of SupTech Report, 2025).
Finally, cognitive limitations play a serious role. As AI-generated research becomes more widespread, consensus is formed more quickly. Chu and Evans (2021) warn that algorithmic systems tend to bolster dominant paradigms, increasing the danger of mental stagnation. If everyone optimizes based on similar data and models, differentiation disappears.
For skilled investors, widespread adoption of AI increases the worth of independent judgment and process diversity as each develop into increasingly rare.
Impact on the investment industry
AI’s growing role in automating investment processes highlights what it cannot eliminate: uncertainty, judgment and accountability. Companies that design their organizations to reflect this reality usually tend to thrive in the approaching decade.
Overall, the evidence suggests that AI is acting as a differentiator somewhat than a universal enabler, widening the gap between firms that embrace reliability, governance and constraints and people who don’t.
At a deeper level, the research points to a philosophical shift. AI’s best value may lie less in prediction and more in reflection—difficult assumptions, surfacing differences of opinion, and forcing higher questions somewhat than simply providing faster answers.
References
Almog, D. AI Recommendations and Non-instrumental Image, Kellogg School of Management Northwestern University, April 2025
by Cast, S. et al. State of SupTech Report 2025, December 2025
Chu, J and J. Evans, Slowed Canonical Progress in Large Fields of Science, October 2021
Gerlich, M., AI tools in society: Implications for cognitive offloading and the long run of critical pondering, 2025
Hendryckx et al. D, A definition of AGI, https://arxiv.org/pdf/2510.18212October 2025
Kalai, A, et al., Why Language Models Hallucinate, 2025, arXiv:2509.04664, 2025
Mahadevan, S. Large causal models from large language models, , https://arxiv.org/abs/2512.07796December 2025
Restrepo, P., We Won’t Miss: Work and Growth within the Age of AGI, July 2025
UC Berkeley, Intesa Sanpaolo, Stanford, IBM Research, measuring equipment in production, , https://arxiv.org/pdf/2512.04123December 2025
