As artificial intelligence becomes increasingly embedded in the Canadian pension ecosystem, the DB pension sector faces a fundamental question: Will AI simply make today’s human-supported pension system more efficient, or could it fundamentally change the role of humans in this system?
For DB pension plans, AI is already influencing both day-to-day operations and longer-term outcomes in several ways:
- Increasing the efficiency of plan administration, such as using AI to review plan member data for accuracy.
- Supporting communication with plan members. For example, AI tools such as expert chatbots can respond directly to member questions.
- As AI becomes more sophisticated, it is increasingly being deployed to support and enhance the services of pension industry providers, such as investment managers, law firms, and retirement consulting firms such as TELUS Health.
- The unprecedented capital investments being made in AI are increasingly influencing plan asset returns, as AI and related technologies affect public market performance and market volatility.
- Finally, the workforce effects of AI may impact a plan’s actuarial liabilities by affecting the rates of hiring of new plan members, rates of member terminations and retirements, and the level of earnings growth of plan members.
The rapid integration of AI into the world of DB pensions raises the question as to how, and even whether, humans will be needed to administer and support DB pension plans in the future.
Given the current state of AI technology, humans clearly continue to play an indispensable role in the world of DB pensions. The Canadian DB pension landscape is extraordinarily complex, with many types of plan designs and multiple sets of legislation applying to pension plans. Actuarial, investment, administration, governance, accounting, legal, and communication expertise are all examples of the broad and deep knowledge needed to effectively oversee and administer a DB pension plan. It is unrealistic to expect that today’s AI tools can accurately navigate all the complexities of today’s DB pension environment without humans playing a primary role.
However, given the unprecedented pace of recent improvements in AI technology, could AI tools become so “smart” in the future that at some point humans will no longer be needed in the DB pension industry?
Imagine a world in which Mary, three months from her 60th birthday, settles into her home office chair and decides that it is time to do some retirement planning. She logs into her employee retirement portal and clicks the "Help me plan for retirement" button displayed on the screen.
A human-like avatar appears. "Hi Mary, I'm Julia, your personal retirement plan representative. I thought you might be reaching out soon. Do you have time for a chat?" Over the next ninety minutes, Julia guides Mary through her DB pension benefits, securely links to her other financial accounts, and builds an individualized and comprehensive retirement picture, including retirement income projections under multiple scenarios.
Reassured by Julia's clear projections and personalized scenarios, Mary realizes that retiring at age 60 is entirely feasible. Julia lays out Mary’s DB pension options, and together they complete every required form during the call. The session ends. Mary takes a deep breath and exclaims, "Wow, I guess I'm retiring!" Then the AI agent gets to work. It first notifies Mary's manager of her decision to retire and updates the HR and payroll systems. It then sends payment instructions to the pension plan's custodian to initiate her pension. Three months later, her first pension cheque is deposited into her account. Not a single human, beyond Mary herself, was involved in this process.
Is a future world as described above inevitable and/or desirable? This question reminds me of an early experience as an actuarial intern in the mid to late 1980s. At the time, the use of spreadsheets was becoming commonplace in pension consulting. Since spreadsheets are flexible and can be programmed to perform actuarial calculations, there was a view that most pension actuaries would no longer be needed in the future because spreadsheets could do the work that most actuaries did.
In retrospect, the opposite was true. Spreadsheets and other improvements in technology did drastically reduce the time spent by actuaries on menial, repetitive tasks. At the same time, they also enabled more complex pension plan designs, more complex and precise actuarial calculations, more customized communications, more timely delivery of actuarial results, and more rigorous pension plan governance. These technology-enabled changes improved the pension system and, in the process, created a lot of work for pension actuaries as well as other industry service providers.
While it is difficult to predict the long-term effects on the DB pension industry of such a transformative technology as AI, I expect that we will experience something similar to the spreadsheets example. AI will undoubtedly continue to increase operational efficiencies in plan administration, freeing up industry professionals to focus on higher-value and customer-facing roles. AI will also further enable customization at the individual level in areas such as member communications, financial planning and possibly even plan design.
AI tools are already supporting pension plan call center staff by enhancing the timeliness, relevance, and quality of information that is available during member calls. As AI tools improve, other areas of member services will likely see improvement, leading to more services being available “on demand” 24 hours a day, seven days a week. Services such as personalized member education as well as financial planning and analysis are just a couple of examples.
Far from removing humans from the equation, much like spreadsheets, AI will allow humans to focus on higher-value tasks in the DB pension environment. This includes creating some roles that previously did not exist at all, such as:
- Understanding and translating the needs and wants of pension stakeholders into practical DB AI tools that address these stakeholder needs more quickly, cheaply, and flexibly than traditional development approaches allow;
- Training and testing the AI tools used by actuarial, administration, and other industry professionals; and
- Providing governance and oversight of the AI technology to ensure it is used ethically and that data privacy, quality, and cybersecurity are maintained. It is important to remember that a plan administrator’s accountability for fulfilling their fiduciary responsibilities to pension plan members cannot be delegated to AI tools.
In a similar vein, it is possible, maybe even likely, that AI will enable humans to innovate in ways that are not even imagined today, leading to opportunities to expand DB or DB-type coverage to more workers and improve the retirement outcomes of Canadians more generally.
Like all models, AI models are fallible. Human oversight will still be needed in the future to interpret model output, apply judgment, provide context and, when appropriate, challenge and override the outputs. Additionally, not everyone will be as willing as Mary was to interact only with an AI agent for a decision as important as deciding on her retirement date.
We should not lose sight of the fact that the primary purpose of DB pensions is to provide secure retirement income to humans. With this primary purpose in mind, as AI technology becomes more powerful and disruptive in the future, my prediction is that the accountability, oversight, and heavy involvement of humans in the DB pension industry will be more important than ever.
|
|
