AINL#010 Augmented Intelligence in Investment Management Newsletter
Welcome to the 010 Edition of the Newsletter on Augmented Intelligence in Investment Management (AINL). Every two weeks, we deliver five unique insights tailored to empower investment decision-makers. Our insights are carefully curated by a seasoned team of market specialists. Unbiased, actionable and practical. They will help you navigate through the noise.
AINL#10 SYNTHESIS
What do these recent developments mean for investment decision-makers?
1. Embrace AI to Enhance Human Decision-Making, Not Replace It
While only 0.01% of UCITS funds currently use AI in formal investment strategies (ESMA, 2025), the practical value lies in supporting—rather than substituting—human expertise. Tools such as LLMs are increasingly used behind the scenes for research synthesis, productivity enhancement, and generating insights (Article 1 & 5). Investment professionals should:
- Experiment with AI tools for informational advantages (e.g. faster synthesis of carbon market data as in Article 5).
- Remain cautious of “AI overreach” by preserving critical human judgment, especially in complex, non-linear, or ambiguous contexts.
- Recognize the first-mover advantage in early but thoughtful adoption of AI-enhanced workflows.
2. Build Behaviorally-Informed, Transparent Decision Architectures
As Cass Sunstein emphasizes (Article 3), nudges offer a way to improve investment decision quality by structuring choices in a liberty-preserving, behaviorally optimized manner. Combined with AI tools, this offers a hybrid path forward:
- Embed behavioral nudges in client-facing tools or internal decision processes to guide toward better outcomes (e.g. setting helpful defaults).
- Use Shapley value explanations (Article 2) to demystify how AI/ML models influence outputs, increasing stakeholder trust and regulatory compliance.
- Ensure decisions are transparent and explainable, bridging AI-driven prediction with human reasoning
3. Evaluate ESG Narratives with Rigor, Not Rhetoric
The DEI debate (Article 4) highlights a broader point: not all widely accepted ESG narratives hold under scrutiny. This has clear implications for investment strategy:
- Prioritize evidence-based assessments of ESG factors, distinguishing between correlation and causation in claims about financial performance.
- Focus on long-term, material impacts (like improved employee retention or regulatory goodwill), not superficial metrics.
- Stay skeptical of unverified ESG claims, particularly from studies lacking peer review or transparency about incentives.
TOP 5 ARTICLES
ARTICLE ONE
Use of AI in Investment Processes.
ARTIFICIAL INTELLIGENCE | ESMA | 3_2025 | Publication > Deck
Important Development
Insightful ESMA report. Their study evaluated the impact of Artificial Intelligence (AI) and its recent advancements in the EU
investment management industry.
First, it studies the operational use of AI by fund managers, i.e. the extent to which the adoption of AI tools by asset managers plays a role in the investment process.
Of the 44.000 Funds screened, only 145 (!) indicated using AI or ML in the investment strategy. Their AuM accumulates to EUR 13 bn or 0.01% of total UCITS funds AuM This number even peaked in 2023. Their use of AI had limited success among investors, with alternating inflows and outflows.
Instead, asset managers use generative AI and tools based on large language models primarily to support human-driven investment decisions, not declared in their formal documentation.
Second, the report assessed investment in AI, i.e. the portfolio allocation to AI-related companies by EU investment funds. Since 2023, actively managed equity funds increased the share of their portfolio invested in a set of AI-driven companies by over 50% – from 9% to 14% – with the market value of these positions doubling.
Why Relevant to You?
The investment industry is at the very beginning of embedding AI and ML in the investment process.
The use of AI for portfolio optimisation is rare (0.01% of all UCITS funds). If adoption is considered, then at company level to enhance productivity and boost capabilities of human decision makers. Altogether, there is still plenty of space for first movers.
ARTICLE TWO
Predicting Financial Market Stress With Machine Learning
ARTIFICIAL INTELLIGENCE | UBIS | Aldasoro et al. | 03_2025 | Paper
Important Findings
A recent BIS working paper proposes novel market condition indicators for the US Treasury, foreign exchange and money market. Like other market condition indicators, the indicators are based on rolling-window principal component analysis. However, the authors argue that the new set of variables used for index construction is better suited to proxy for volatility, illiquidity, and arbitrage breakdowns.
Moreover, the new indicators appear to be less sensitive to the VIX than other financial conditions indicators, which results in better sensitivity to market-specific stress episodes. The constructed indicators are then used as endogenous variables in a forecasting exercise with different quantile regression approaches. The authors show that a tree-based approach tends to have a better out-of-sample performance than univariate and multivariate quantile regression, attributing this finding to the ability of ML models to better deal with non-linear dynamics and sparse signals. Finally, Sharpley values (SHAP) are used to attribute model predictions to individual explanatory variables.
Why Relevant to You?
Episodes of market stress can severely impair the efficient allocation of capital and price formation, leading to broader economic disruptions. The development of robust measures to gauge market conditions is thus crucial for policymakers and market participants alike. The new market indicators can be used as real-time diagnostic tools, flagging market-specific stress and potentially posing an arbitrage opportunity.
With respect to the ML regression approach, the authors argue that it could serve as a template for stress-testing frameworks for tail risk estimation. Finally, as both investors and regulators stress the importance of explainability for model-based investment and policy decisions, Shapley values could be one way to ensure that humans can understand how complex models arrive at their output.
ARTICLE THREE
Nudges and Nudging: A User’s Manual
HUMAN & ARTIFICIAL INTELLIGENCE | Harvard University | Cass Sunstein | 03_2025 | Paper
Important Findings
Cass Sunstein is one of the still living pillars in behavioral sciences. His research on nudges and sludges together with Richard Thaler were breakthroughs in making behavioral considerations applicable. In his recently published paper he thoroughly examines various aspects of nudges, starting with their definition as liberty-preserving techniques that guide people in certain directions while allowing them the freedom to choose.
Why Relevant to You?
In addition to defending their utility from misconceptions, he offers a practical user’s manual on how to embed nudges in a decision design. Investment decision designs are not only lagging in their adaptivity of artificial forms of intelligence – see article above – but also in behaviorally optimized elements. Nudges can be one of those elements to trigger outcomes of higher decision quality. Of utmost relevance for all of us.
ARTICLE FOUR
DEI Can Be Justified, But Not by Better Profits
HUMAN & ARTIFICIAL INTELLIGENCE | Bloomberg Opinion | 2_2025 | Article
Important Findings
Companies claim that diversity, equity, and inclusion (DEI) drive better business results—but what if that’s not true? While industry studies promote DEI’s financial benefits, deeper research tells a different story. Does hiring for diversity actually improve profits, or is it just a well-intentioned myth? Allison Schrager unpacks the real impact of DEI on business performance.
Why Relevant to You?
Investors should be cautious about claims that DEI boosts profitability, as rigorous research finds no clear link between diversity and financial performance. Many pro-DEI studies lack peer review or come from firms with a financial stake in promoting it. While diversity of thought can enhance decision-making, demographic diversity alone doesn’t guarantee better outcomes. However, DEI can improve workplace culture and employee satisfaction, which may indirectly support long-term stability. Investors should evaluate DEI policies based on their real impact, not just corporate narratives.
ARTICLE FIVE
GREEN IQ – A Deep Search Platform for Comprehensive Carbon Market Analysis
HUMAN & ARTIFICIAL INTELLIGENCE | Oluwole Fagbohun et al | 03_2025 | Paper
Important Findings
A recent study showed promising results for an AI-powered deep search platform designed to make carbon market intelligence less labor-intensive. Compared to traditional approaches, significant time savings and cost-reduction have been achieved through LLM-powered autonomous analysis and automated report generation (digesting vast amounts of heterogeneous data from policy documents, industry reports, academic literature, and real-time trading platforms).
Why Relevant to You?
If confirmed, such approaches/tools come with a double take-away. On the one hand, it will offer an efficient framework for environmental and financial intelligence, enabling more accurate, more timely, and cost-effective decision-making in complex regulatory landscapes. On the other hand, it increases the importance of getting professionals human-proof when using technology. There is a risk to loose grip on the underlying market reality if it gets topped with a Gen-AI layer (eg. to speed up decision taking even more).
As indicated in our AINL_008 edition, people overestimate the degree of alignment between GenAI’s choices and human choices. It is likely to assume that this risk gets further increased if one has neither really dug through data, nor has critically challenged decision proposals. One of them will remain key to ensure impact in complex situations, even/especially with high-speed AI tools.
