While emerging forms of artificial intelligence such as agentic AI may hold promises and challenges for pension plan sponsors, the human touch remains important when advising plan members on financial decisions, said Francis Boulianne, principal in Normandin Beaudry’s group benefits practice, during a recent webinar hosted by the consultancy.
“[Chatbots] were never intended to replace financial guidance or advice or any sort of financial planning, nor was it capable of doing so. What we’re moving into today is this new reinvigorated promise that agentic AI will be able to someday provide that [and] we are in the middle of widespread experimentation [with the technology]. . . . Thankfully, [the majority of our plan sponsor clients] are using it to try to fill the gaps . . . from an administration perspective.”
Read: AI threatening to push young workers out of entry-level jobs
Currently, he noted the use of AI in pension plans is focused in the defined benefit space, in which clients are using the technology to calculate service buy-backs and the amount of money plan members will need to save in order to retire at their desired age.
Legislation and pension governance practices will also need to catch up to the technology. “I think the first step [is] understanding where our fiduciary liability starts and stops [and] understanding what sort of data we’re sharing that our third-party providers are utilizing.”
Also speaking during the webinar, Dimitri Poliak, a principal at Normandin Beaudry, said in the benefits space, AI is being used by plan sponsors for underwriting as well as fraud detection and plan personalization.
Similar to pensions, the technology isn’t a panacea to plan sponsors’ and members’ benefits concerns. “We have a client that wanted to use AI to make [benefits] recommendations to employees. The short answer is yes, they can do that, but as a best practice, we don’t even recommend employers to make recommendations about which option employees should select.”
Read: Employers urged to close AI adoption gap for women: report
Even seemingly simple tasks such as creating job postings still require a human element to ensure they’re completed correctly, said Darcy Clark, a senior principal at Normandin Beaudry.
“For example, you can pump out a job posting in 30 seconds [with AI] but does it have the skill sets the business is looking for? There’s a lot of risks and dangers of just relying on AI technology to create postings.”
And as compensation teams shrink and create a knowledge gap, the risks of AI increase accordingly. “Compensation teams are getting leaner, so we’re wearing a lot of different hats and losing that skill set and AI can’t just replace the nuances in the art of compensation. . . . That still belongs to us.”
Read: Quebec’s AI workplace guidelines signal a shift for employers: expert