The benefits of artificial intelligence go hand in hand with the many risks, but the pension industry’s requirements around fiduciary responsibility and governance, as well as privacy concerns, add an important layer to these considerations.
Genevieve Hayman, senior affiliate researcher in research, advocacy and standards, the CFA Institute
Many artificial intelligence risks — which include data privacy and security concerns, lack of transparency and accountability issues — aren’t exclusive to pensions.
However, pension funds face distinct AI challenges when it comes to forecasting, member engagement and governance. Since pensions are oriented for the long term, historical data may not accurately predict future trends in areas like longevity, interest rates or alternative assets. Determining appropriate levels of model fit (not overfitting or underfitting to historical data) will be a significant challenge. Data that may be representative of a current population, economy or regulatory environment may be inappropriate for forecasting 20 or 30 years into the future.
Another area where pensions must be cautious is in member personalization and engagement. AI-driven robo-advisors and dashboards can offer individually curated suggestions, but too much automation could decrease financial literacy, potentially leading members to follow AI-generated recommendations without understanding their financial situation.
On the other hand, if members are provided with flexibility and choice, they’re likely to be overwhelmed and decision avoidant. Developing AI-driven products that personalize, while fostering appropriate levels of engagement, is critical in the design process.
Finally, pensions are subject to strict fiduciary and governance obligations, meaning trustees and plan sponsors must act in the best interests of members and be able to justify investment decisions. Technology providers may be unaware of the exact nature of these obligations or their products may be designed without sufficient guardrails for meeting a fiduciary standard.
Read: Effective AI integration requires customization effort from pension plan sponsors: report
While the CFA Institute’s report, Pensions in the Age of Artificial Intelligence, details many AI uses in pension governance, technology can’t replace the decision-making capacity of trustees who understand the needs, goals and values of the pension fund. AI can be a helpful tool, but pension administrators still have to get the fundamentals right.
Cathy Cobey, global AI responsible leader, assurance, EY
Using AI in pension administration and governance offers numerous benefits, including increased efficiency and automation, advanced data management, predictive analytics and potential cost savings. There are also underlying risks in using AI and machine learning, with some of these inherent across industries, including bias, data privacy and security considerations and regulatory and compliance challenges.
In the specific context of pensions, risks include: erroneous benefit calculations — AI-driven automation in administration can lead to incorrect pension payouts if algorithms misinterpret data or apply incorrect assumptions; longevity and financial modeling risks — AI models predicting life expectancy or investment returns might produce inaccurate forecasts, affecting pension fund sustainability; and misinterpretation of complex regulations — pension policies and regulatory frameworks are intricate and frequently changing, so AI systems may struggle to interpret nuanced legal requirements, leading to compliance risks.
Read: AI holds uncapped potential for pension plan sponsors’ administration, communications: expert
Companies can think about safe AI adoption by integrating ethical principles and robust governance frameworks into their AI strategies. This involves ensuring accountability by assigning clear ownership and responsibility for AI systems throughout their lifecycle. Companies should prioritize explainability, making AI decision-making processes transparent and understandable to appropriate stakeholders.
Compliance with relevant laws, regulations and professional standards is essential, as is implementing stringent data protection measures to safeguard sensitive information. Also, engaging with stakeholders and staying informed about emerging ethical, social and regulatory issues will further support the responsible and safe adoption of AI technologies.
While AI offers efficiency and predictive power, careful governance, continuous auditing and human oversight are crucial to mitigate these risks and build trust and confidence in AI used in pension administration.
Read: How Texas’ Teacher Retirement System is incorporating AI into pension governance