'Human intelligence must be the ultimate deciding factor,' highlights CFA Institute report.
AI is steadily shaping the world of pension management, offering new efficiencies, deeper insights, and consistent challenges for pension managers, according to recent research from the CFA Institute.
Genevieve Hayman, senior research affiliate at the CFA Institute, explained while AI’s role in pension management is still developing and surpassing efficiency gains, it’s not yet replacing the human decision-making process. Rather, machine learning tools are enhancing their ability to be more informed and accountable in their fiduciary duties, with AI being more of a supportive tool.
“Pension trustees have a very complex set of responsibilities, so you see AI being implemented in everything from trustee decision support, information gathering tools, to investment manager analytics and insights to stewardship benchmarking,” said Hayman.
Hayman emphasized one of the key findings outlined in CFA’s report Pensions in the Age of Artificial Intelligence is AI’s growing role in pension governance. Traditionally, trustees have relied on manual research and financial models to guide their decision-making whereas now, AI is being used to mitigate this.
“Although AI is useful in the decision-making process, human intelligence must be the ultimate deciding factor,” said the report. “For AI to be successfully implemented within the pensions value chain, particularly in governance and investment processes, the AI must be explainable and transparent. It is vital that trustees be able to justify the reasoning behind any suggestion they make.”
Risk management is a natural concern when integrating AI into pension management. Additionally, AI is playing a growing role in de-risking strategies. Hayman highlighted a case study involving Cardano’s Discover AI tool, which focuses on employer covenant assessments in the UK market. In the UK, “covenant” is the support provided by a plan sponsor to a plan.
“One of the things AI is really helping with is being responsive to data in a more real-time way,” said Hayman. “How can we quickly get an overview of our pension funds and get a snapshot overview of how they're performing?” she noted, adding machine-learning tools allow pension managers to track funding levels and market scenarios more efficiently.
“You want people to be making good decisions for their retirement, to be engaging with their plan regularly and that's been a struggle in a historical context,” added Rhodri Preece, senior head of research at CFA Institute.
“With the advent of these new tools available, it's a way in which we can better engage our audience and get them thinking more proactively about saving for their retirement.”
Another major development is AI’s ability to personalize the pension experience for members. Historically, pension plans have relied on the “one-size-fits-all” approach, but AI is making it possible to tailor recommendations based on individual financial situations, Hayman noted.
Beyond governance, AI is making an impact in asset liability management. While Hayman cautions it may not revolutionize the entire approach, it is enhancing “how we gather information that goes into actuarial forecasts... to make better and more responsive decisions based on changing climates and environments,” she said.
Investment management is another area where AI is proving useful, although its applications aren't exclusive to pensions. AI-driven tools are being used for investment analysis, portfolio optimization, and risk management and its ability to process large, unstructured data sets is helping pension managers identify new opportunities and assess risks with greater accuracy.
Hayman points to Japan’s Government Pension Investment Fund (GPIF), the world’s largest pension fund, who partnered with Sony to enhance its investment manager selection process. AI-driven accountability measures help trustees assess whether managers are performing up to standard and justify investment decisions, said Hayman.
Hayman also highlighted Australia’s State Super, which has developed a reinforcement learning tool to balance short- and long-term investment strategies.
“It’s always a balance with pension funds because you're talking about long time horizons but also data that's becoming available with each passing moment. The learning tool can take data on an ongoing basis about current markets and then use that to train their longer time horizon forecasting,” Hayman explained.
Despite AI’s advantages, the fear of automation replacing human expertise continues to be one of the most common. However, Hayman emphasized that AI is not replacing trustees or investment managers, it’s only enhancing their ability to make informed decisions.
“The decisions are still being made with those fiduciary individuals, but they’re tools that help make those decisions better,” she said.
Preece agreed, adding that a human is always needed “in the loop to ensure the accuracy and the validity of any decision processes that comes from using AI. I don't foresee that changing.”
When it comes to investment returns, Preece is careful not to make sweeping claims about AI’s ability to outperform human-driven strategies.
“If the question is, does an AI-driven approach deliver superior returns to a human-only driven approach, then that's impossible to answer,” he said. “It depends on your strategy, it depends on your time horizon, it depends on your liquidity constraints. There's any number of factors that will determine the success of any given strategy, including an AI-driven one.”
While AI is becoming a valuable tool for pension managers, its success ultimately depends on strategic implementation.
“We’re still in the relatively early stages of AI adoption in financial services, but there’s some clear pointers. If you incorporate these tools into your processes, you can ultimately deliver better value for clients,” Preece noted.
