Artificial intelligence is transforming how investment decisions are made, and it is here to stay. Used wisely, it can sharpen professional judgment and improve investment outcomes. But the technology also carries risks: today’s reasoning models are still underdeveloped, regulatory guardrails are not yet in place, and overreliance on AI outputs could distort markets with false signals.
This post is the second installment of a quarterly reflection on the latest developments in AI for investment management professionals. It incorporates insights from a team of investment specialists, academics, and regulators who are collaborating on a bi-monthly newsletter for finance professionals, “Augmented Intelligence in Investment Management.” The first post in this series set the stage by introducing AI’s promise and pitfalls for investment managers, while this post pushes further into risk frontiers.
By examining recent research and industry trends, we aim to equip you with practical applications for navigating this evolving landscape.
Practical Applications
Lesson #1: Human + Machine: A Stronger Formula for Decision Quality
The fusion of human and machine intelligence strengthens consistency, which is a key marker of decision quality. As Karim Lakhani of Harvard Business School summarized: “It’s not about AI replacing analysts—it’s about analysts who use AI replacing those who don’t.”
Practical Implication: Investment teams should design workflows where human intuition is complemented, not replaced, by AI-driven reasoning aids, ensuring more stable decision outcomes.
Lesson #2: Humans Still Own the Uncertainty Frontier
Current limitations of large reasoning models (LRM), which can think through a problem and create calculated solutions, mean it is up to investment managers to decipher the impact of less structured imperfect markets. Frontier reasoning models collapse under high complexity, reinforcing that AI in its current form remains a pattern‑recognition tool.
While the new generation of reasoning models promise marginal performance improvements such as better data processing or forecasting, the results do not live up to the promises. In fact, the less structured a market phenomenon, the more failure-prone the models’ outcomes.
Practical Implication: Transparency around benchmark sensitivity and prompt design is vital for consistent use in investment research.
Lesson #3: Regulators Enter the AI Arena
Supervisory authorities are piloting Generative AI (GenAI) for process automation and risk monitoring, offering case studies for industry adoption. Regulators are quickly identifying a bevy of vulnerabilities pertaining to AI that could negatively impact financial stability. A report issued by the Financial Stability Board (FSB) which was established after the 2008 financial crisis to promote transparency in financial markets, pointed out a number of potential negative implications. GenAI can be used to spread disinformation in financial markets, the group said. Other possible issues include third-party dependencies and service provider concentration, increased market correlation due to the widespread use of common AI models, and model risks, including opaque data quality. Cybersecurity risks and AI governance were also on the FSB’s list.
To wit, regulators are on alert, working on their own integration of AI applications to address the systemic risks explored.
Practical Implication: Adaptive regulatory frameworks will shape AI’s role in financial stability and fiduciary accountability.
Lesson #4: GenAI as a Crutch: Guarding Against Skill Atrophy
GenAI can boost efficiency, particularly for less-experienced workers, but it also raises concerns about metacognitive laziness, or the tendency to offload critical thinking to a machine/AI, and skill atrophy. Structured AI‑human workflows and learning interventions are critical to preserving deep industry engagement and expertise.
GenAI firm Anthropic’s analysis of student AI use shows a growing trend of outsourcing high-order thinking, like analysis and creation, to GenAI. For investment professionals, this is a double-edged sword. While it can boost productivity, it also risks atrophy of core cognitive skills critical for contrarian thinking, probabilistic reasoning, and variant perception.
Practical Implication: Investors must ensure that AI tools do not become a crutch. Instead, they should be embedded in structured decision-making and workflows that preserve and even sharpen human judgment. In this new environment, developing metacognitive awareness and fostering intellectual humility may be just as valuable as mastering a financial model. Investing in AI literacy and piloting AI‑human workflows that preserve critical human judgment will serve to foster and perhaps amplify, cognitive engagement.
Lesson #5: The AI Herd Effect Is Real
Being contrarian in seeking alpha means understanding the models everyone else is using. Widespread use of similar AI models introduces systemic risk: increased market correlation, third-party concentration, and model opacity.
Practical Implication: Investment professionals should:
Diversify model sources and maintain independent analytic capabilities.
Build AI governance frameworks to monitor data quality, model assumptions, and alignment with fiduciary principles.
Stay alert to information distortion risks, especially through AI-generated content in public financial discourse.
Use AI as a thinking partner, not a shortcut—build prompts, frameworks, and tools that stimulate reflection and hypothesis testing.
Train teams to challenge AI outputs through scenario analysis and domain-specific judgment.
Design workflows that combine machine efficiency with human intent, especially in investment research and portfolio construction.
Conclusion: Navigate the AI Risk Frontier with Clarity
Investment professionals cannot rely on the overly confident promises made by artificial intelligence firms, whether they come from LLM providers or related AI agents. As use cases grow, navigating emerging risk frontiers with mindfulness of what they can and cannot add in improving the investment decision quality are of paramount importance.
Appendix & Citations:
Fagbohun, O., Yashwanth, S., Akintola, A. S., Wurola, I., Shittu, L., Inyang, A., . . . Akinbolaji, T. (2025). GreenIQ: A deep search platform for comprehensive carbon market analysis and automated report generation. arXiv.
Handa, K., Bent, D., Tamkin, A., McCain, ., Durmus, ., Stern, M., . . . Ganguli, D. (2025, April 8). Anthropic Education Report: How university students use Claude. Retrieved from Anthropic: https://www.anthropic.com/news/anthropic-education-report-how-university-students-use-claude
van Zanten, J. (2025). Measuring Companies’ Environmental and Social Impacts: An Analysis of ESG Ratings and SDG Scores. Organization and Environment.
Brynjolfsson, E., Li, D., & Raymond, L. (2025). Generative AI at Work. The Quarterly Journal of Economics.
Pérez‑Cruz, F., & Shin, H. (2025). Putting AI agents through their paces on general tasks. Bank for International Settlements (BIS).
Ren, Y., Deng, X. (., & Joshi, K. (2024). Unpacking Human and AI Complementarity: Insights from Recent Works. SSRN.
Traub, B., Traub, I., Peper, P., Oravec, J., & Thurman, P. (2023). Modeling the AI-Driven Age of Abundance: Applying the Human-to-AI Leverage Ratio (HAILR) to Knowledge Work. SSRN Electronic Journal.
Schmälzle, R., Lim, S., Du, Y., & Bente, G. (2025). The Art of Audience Engagement: LLM‑Based Thin‑Slicing of Scientific Talks. arXiv.
Otis, N., Clarke, R., Delecourt, S., Holtz, D., & Koning, R. (2023). The Uneven Impact of Generative AI on Entrepreneurial Performance. OSF Preprints.
Fan, Y., Tang, L., Le, H., Shen, K., Tan, S., Zhao, Y., . . . Gašević, D. (2024). Beware of Metacognitive Laziness: Effects of Generative Artificial Intelligence on Learning Motivation, Processes, and Performance. British Journal of Educational Technology.
Financial Stability Board. (2024). The Financial Stability Implications of Artificial Intelligence. Financial Stability Board.
Financial Policy Committee, Bank of England. (2025). Financial Stability in Focus: Artificial Intelligence in the Financial System. Bank of England.
Qin, Y., Lee, R., & Sajda, P. (2025). Perception of an AI Teammate in an Embodied Control Task Affects Team Performance, Reflected in Human Teammates’ Behaviors and Physiological Responses. arXiv.
Gao, K., & Zamanpour, A. (2024). How can AI‑integrated applications affect the financial engineers’ psychological safety and work‑life balance: Chinese and Iranian financial engineers and administrators’ perspectives. BMC Psychology.
Backlund, A., & Petersson, L. (2025). Vending‑Bench: A Benchmark for Long‑Term Coherence of Autonomous Agents. arXiv.
Xu, F., Hao, Q., Zong, Z., Wang, J., Zhang, Y., Wang, J., . . . Gao, C. (2025). Towards Large Reasoning Models: A Survey of Reinforced Reasoning with Large Language Models. arXiv.
Daly, C. (2025, May 8). Klarna Slows AI‑Driven Job Cuts With Call for Real People. Retrieved from Bloomberg: https://www.bloomberg.com/news/articles/2025-05-08/klarna-turns-from-ai-to-real-person-customer-service?embedded-checkout=true
Hämäläinen, M. (2025). On Psychology of AI – Does Primacy Effect Affect ChatGPT and Other LLMs? arXiv.
Schmälzle, R., Lim, S., Du, Y., & Bente, G. (2025). The Art of Audience Engagement: LLM-Based Thin-Slicing of Scientific Talks. arXiv.
Bednarski, M. (2025, May–June). Why CEOs Should Think Twice Before Using AI to Write Messages. Harvard Business Review.
Shojaee, P., Mirzadeh, I., Alizadeh, K., Horton, M., Bengio, S., & Farajtabar, M. (2025). The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity . Apple Machine Learning Research, Apple Inc.
Meincke, L., Mollick, E., Mollick, L., & Shapiro, D. (2025). Prompting Science Report 1: Prompt Engineering is Complicated and Contingent. Generative AI Labs, The Wharton School of Business, University of Pennsylvania.
Ivcevic, Z., & Grandinetti, M. (2024). Artificial intelligence as a tool for creativity. Journal of Creativity.
Zhang, J., Hu, S., Lu, C., Lange, R., & Clune, J. (2025). Darwin Gödel Machine: Open‑Ended Evolution of Self‑Improving Agents. arXiv.
Foucault, T., Gambacorta, L., Jiang, W., & Vives, X. (2024). Artificial Intelligence in Finance. Centre for Economic Policy Research (CEPR).
Kosmyna, N., Hauptmann, H., Yuan, Y., Situ, J., Liao, X.‑H., Beresnitzky, A., . . . Maes, P. (2025). Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task. arXiv.
Vasileiou, S., Rago, A., Martinez, M., & Yeoh, W. (2025). How Do People Revise Inconsistent Beliefs? Examining Belief Revision in Humans with User Studies. arXiv.
Prenio, J. (2025). Starting with the basics: A stocktake of gen AI applications in supervision. Financial Stability Institute, Bank for International Settlements (BIS).
