Viewpoints

Sub-Adviser AI Oversight: Focus on what matters

May 15, 2025

By Joyce Li, CFA

Comprehensive guidance on the oversight of sub-advisers that exists in the market has served mutual fund boards well for over 15 years. These frameworks remain foundational, yet the widespread adoption and increasing sophistication of artificial intelligence (AI) and related technologies now embedded in investment processes, operational functions, and compliance systems demand thoughtful adaptation of the oversight approach. Indeed, AI adoption and implementation is transforming how investment advisers and sub-advisers operate, and these technological advances can pose distinct oversight challenges for independent fund directors. As this transformation gains momentum, the need for updated guidance—and frequency of challenges—will intensify.

 

According to Mercer's 2024 global manager survey of 150 asset managers, 54% of investment managers already use AI, with another 37% planning adoption [1]. While many firms leverage AI for operational efficiency in areas like client reporting, compliance monitoring, and trade execution, adoption in investment processes remains more selective. Managers report using AI for alternative data analysis, idea generation, and developing forward-looking indicators, though most investment teams maintain significant human oversight. Private market investment advisers, in particular, have been exploring sophisticated AI applications for deal sourcing, due diligence, and valuation.

 

For fund boards overseeing multiple sub-advisers through a principal adviser, AI's inherent pace of change creates a fundamental oversight challenge, making it difficult to determine which changes materially affect a sub-adviser's ability to deliver on their mandate.

 

It is important to establish scope for our discussion. Fund boards aren't responsible for overseeing every AI experiment at a sub-adviser's firm. Rather, the board’s oversight should focus on AI applications that materially affect the sub-adviser's ability to execute their investment mandate, investment performance and risk management, conflicts of interest that could disadvantage the fund, and/or business model changes that redirect resources away from the fund. Other applications, while interesting, may not warrant board attention. 

 

Two Key Oversight Domains

Fund boards need different oversight approaches depending on where and how sub-advisers deploy AI. While AI's integration into investment management continues to evolve, boards can focus their attention on two distinct areas: operational applications and investment process integration.

 

When AI Stays in Operations: Most investment managers leverage AI for operational efficiency, with applications in client reporting, compliance monitoring, trade execution, and other areas. For instance, BlackRock's Aladdin Copilot facilitates faster user onboarding and generates reports, while Franklin Templeton collaborates with Microsoft to embed AI in sales, marketing, and client support functions.

 

These operational transformations bring tremendous business value but also introduce layers of risk that fund boards must now consider. In April, MFDF and Deloitte collaborated on a white paper, “Role of the Fund Director in the Oversight of the Risk Management Function,” that emphasizes AI's implications for data privacy, cybersecurity, and operational resilience [2]. When sub-advisers process proprietary or third-party fund data through AI systems, boards should understand how sensitive information is protected and whether data governance practices meet regulatory requirements.

 

The competitive landscape adds urgency. As more investment managers adopt AI for operational efficiency, sub-advisers face pressure to keep pace or risk falling behind in service quality and cost effectiveness. This creates a delicate balance: Rushing AI adoption without proper controls could introduce operational risks that might ultimately harm fund performance or investor protection. Third-party AI dependency creates additional risks; if critical operations rely on external platforms, boards should understand contingency plans for vendor disruptions or system failures.

 

For compliance functions, AI's influence grows more complex. While AI can enhance surveillance and monitoring, it also requires new oversight mechanisms. Boards should understand how sub-advisers validate AI-driven compliance systems and what safeguards exist against algorithmic errors or biases that could lead to regulatory violations.

Additionally, when AI automates routine processes, does the sub-adviser reinvest efficiency gains into investment capabilities, or do these savings simply enhance their margins? Are any benefits passed through to fund investors?

 

The core oversight question is this: Does operational AI strengthen the sub-adviser's ability to focus on its primary mandate, or does it introduce complexity that could detract from its core responsibilities to the fund?

 

When AI Enters Investment Processes: A growing cohort of sub-advisers integrate AI directly into investment processes. These applications span model development, alternative data analysis, portfolio construction, and risk management. The Mercer survey reveals that 53% of managers use AI for individual security selection, while 43% incorporate alternative datasets.

 

This integration requires boards to revisit fundamental questions from the original sub-advisory due diligence. When the principal adviser first recommended this sub-adviser, what specific expertise was being accessed? If that expertise is now augmented or replaced by AI capabilities, does the original rationale still hold? Independent Directors Council 2010 guidance on oversight of sub-advisers [3] emphasizes understanding "the reasons why a fund's adviser recommends the use of a sub-adviser." AI adoption may impact the sub-adviser's distinctive advantage or fundamentally alter its approach. Key questions for boards to consider include: Can the sub-adviser explain AI's impact on investment decisions in terms the board understands? Does integration create new regulatory compliance risks, particularly around fair disclosure or best execution?

 

Performance monitoring takes on new dimensions when AI is introduced. MFDF guidance on sub-adviser oversight from 2009 highlights the board's responsibility to oversee "investment performance and the effectiveness of its compliance program.”[4] AI-driven investment processes may require enhanced performance attribution analysis, and boards should monitor whether AI models create unintended biases in security selection or contribute to style drift that conventional analysis might miss. Consider asking: Is there evidence of "model overfitting" where AI performs well on historical data but struggles with new market conditions? How does the sub-adviser detect and correct for algorithmic bias? What happens when AI models provided by third-party vendors continue to evolve?

 

When AI is involved, risk management requires enhanced scrutiny. Directors should understand how sub-advisers validate AI models and manage risks like model drift, data quality issues, and AI hallucinations. Critical considerations include data privacy and security, especially when using alternative datasets containing sensitive information. Consider asking: What happens when primary data sources become unavailable? How does the sub-adviser protect against adversarial attacks? Are there adequate controls to prevent unauthorized model changes that could affect investment decisions?

 

Lastly, AI adoption requires additional efforts when it comes to regulatory compliance and disclosure. When reviewing sub-advisory agreements, boards should ensure the sub-adviser has adequate resources to maintain regulatory compliance as AI capabilities evolve. Are there controls to prevent AI systems from acting on material nonpublic information embedded in alternative data? How transparent are AI-driven investment decisions in client reporting and regulatory filings?

 

The critical evaluation: Is AI enhancing the specific investment expertise that was originally engaged, or has it fundamentally altered the investment? This assessment should guide decisions about continuing or terminating the sub-advisory relationship based on fund shareholders' best interests.

 

Hidden Conflicts of Interest

AI introduces new potential conflicts of interest that traditional oversight frameworks might miss. Here are two examples:

 

Consider cost allocation conflicts. When a sub-adviser develops proprietary AI capabilities, who bears the development costs versus who reaps the benefits? If the fund's fees help finance AI development that primarily enhances the sub-adviser's competitive position or benefits the sub-adviser’s proprietary funds, this creates a conflict. Boards should understand how AI development costs are allocated and whether the benefits flow proportionally to all clients.

 

Data usage presents another subtle conflict. Sub-advisers may use aggregated trading data, portfolio holdings information, or market insights from managing the fund to train AI models. While this might improve the sub-adviser’s overall capabilities, it raises questions about who owns the intellectual property created from fund data. If these AI models later benefit other clients or the sub-adviser's proprietary strategies, the fund has effectively subsidized the sub-adviser’s broader business development. The IDC framework notes that boards should receive "a complete picture" of relationships with sub-advisers, and this now includes understanding how fund data feeds into AI systems.

 

Implementing Adaptive Oversight

The AI oversight challenge will only intensify. AI capabilities are advancing from pattern recognition to sophisticated reasoning. Privacy regulations increasingly clash with AI's appetite for data, while competitive pressures drive rapid adoption without adequate testing. As talent wars heat up, resources that once supported investment teams may be redirected to AI specialists.

 

Fund boards must maintain effective oversight without getting into the weeds. The answer lies in adapting existing frameworks with targeted refinements.

 

First, establish materiality thresholds that reflect a fund's specific circumstances. Work with the principal adviser to define clear triggers for board notification, such as when AI enters investment decisions, significant resources shift to AI initiatives, or new critical vendor dependencies emerge.

 

Second, leverage existing reporting channels effectively. Principal advisers already filter information for boards; ensure the adviser has the expertise to evaluate AI developments appropriately. This might mean the adviser must enhance its technology due diligence capabilities or engage specialized consultants for periodic assessments.

 

Third, integrate AI considerations into standard review processes. During annual sub-advisory agreement evaluations, examine whether AI adoption has altered the original value proposition. When reviewing performance, look for patterns that might indicate algorithmic influence versus human judgment.

 

In addition, consider these strategic questions:

 

  • Is the sub-adviser's AI adoption consistent with similar managers in its peer group?
  • Have compliance and risk management systems kept pace with AI integration?
  • Do we need independent verification of AI-related claims or capabilities?
  • Are AI efficiency gains benefiting fund shareholders or just improving the sub-adviser's margins?

 

The MFDF guidance emphasizes termination provisions when relationships no longer serve shareholder interests. For example, if a quantitative manager hired for proprietary models now relies heavily on third-party AI platforms, this might fundamentally alter its value proposition and warrant board action.

 

Focus on What Matters

Fund boards don't need to become AI experts to maintain effective oversight. Directors’ fiduciary responsibility remains unchanged: Ensure sub-advisers deliver the expertise shareholders expect with appropriate risk controls.

 

The frameworks that have served boards well for decades remain valid. Focus on outcomes rather than technical details. Ask whether AI changes enhance or hinder a sub-adviser's ability to serve the fund's shareholders.

 

With clear materiality thresholds, strategic use of existing oversight channels, and attention to what truly impacts shareholder interests, boards can navigate this technological transition while staying true to their core mission.


Joyce Li, CFA, is CEO of Averanda Partners, where she advises business leaders at the intersection of AI, finance, and governance. She has nearly two decades of investment management experience, including as a mutual fund portfolio manager at Matthews International Capital Management. An experienced board director, Li serves on the advisory board of the technology startup OpenBB and previously served as president on the CFA Society San Francisco board and on other nonprofit boards. Li co-authored the Athena Alliance AI Governance Playbook and created the “AI Simplified for Leaders” newsletter with curated strategic insights.


[1] "AI integration in Investment Management," a global manager survey by Mercer Investments, March 2024

[2] "Role of the Fund Director in the Oversight of the Risk Management Function," a  paper by Deloitte and MFDF, April 2025

[3] "Board Oversight of Sub-Advisers," a task force report by Independent Directors Council, January 2010

[4] "Practical Guidance for Directors on the Oversight of Sub-Advisers," a report by Mutual Fund Directors Forum, April 2009

 

 

Most Read

10 Things
10 Things...to know about the ETF share class

Independent directors who anticipate overseeing exchange-traded fund share classes are contemplating myriad issues related to what many industry participants believe will be a seismic shift in ...