Retirement savings and behavioral finance expert Shlomo Benartzi is turning his attention to artificial intelligence.

In a recent column in The Wall Street Journal, he laid out the ways he sees artificial intelligence aiding in financial planning and advisement. He also discussed some of its problems, such as the technology’s potential to give educated, but ultimately wrong, answers and its propensity to “double down” after a bad bet to try and recoup losses.

“A lot of improvements are needed before AI can become an effective financial adviser,” Benartzi wrote. “Nevertheless, it’s clear that AI will play an important role in the future of financial advice.”

The UCLA professor emeritus and founder of PensionPlus is also starting to explore AI’s role in the qualified retirement plan space, he tells PLANADVISER.“You can think about the plan participant, and what I think AI does is give access to personalized advice,” he says.

Benartzi is working on a project to see if AI can help personalize content for plan participants at scale—though it is in the early stages.

“Can the machine help take the same idea about the need to save and tell me the style I need to write it in for people of different income, people of different gender, people of different whatever?” he asks. “I think it would allow us to do 80% of the assessment with 20% of the effort.”

The Human Touch

Financial wellness programs provider Financial Finesse is already operating in a hybrid AI and human coaching environment, according to Liz Davidson, its founder and CEO.

The firm’s AI-driven service, Aimee, walks employees through a personalized path that will respond to financial concerns and questions. If situations warrant human interaction, it will pass the user to a live financial coach with Certified Financial Planner certification. The company’s service was recently added by Capital Group and is in use by plan sponsors and retirement solutions providers, according to the firm.

“I call it mass personalization,” Davidson says. “You can reach so many people, but in such a deep way. And that was not possible before.”

The coaching, or human, element is still crucial, Davidson says, because it personalizes guidance in a way that the “technology cannot do for most people.”

She also notes that the information provided to participants is drawn from a discrete pool curated by Financial Finesse. Unlike open architecture AI tools that draw from the overall internet, responses are limited to the firm’s internal knowledge base.

Even when operating in a discrete system, however, it takes diligence and testing to ensure the technology is not “hallucinating” answers. Financial Finesse has a team of people monitoring and adjusting for responses, according to the firm. There, again, the human element is crucial in leveraging AI’s capabilities.

Proceed Carefully

Mercer and the CFA Institute dedicated a section of their recent global pension report to AI’s benefits and potential dangers in the retirement plan space.

On the positive side, the organizations’ report also discussed the technology’s ability to further both retirement plan management and participant outcomes.

“There is a need for the development of AI tools to provide members with meaningful options and recommendations for … complex decisions,” the researchers wrote. “This is likely to be an area where AI could be used to provide a more affordable and realistic outcome than expecting every retiree to see a personal financial advisor.”

Mercer and the CFA also pointed out, however, potential problems. Operationally, AI may struggle with true personalization. There is potential for participants to get incorrect information or be guided to a solution that—while good for a similar situation—may not really fit their own, according to the report.

The researchers also pointed out the potential for retirement plan savings fraud stemming from the technology. AI, the researchers noted, can “faithfully reproduce a person’s voice, writing style, photo or video.”

“This, combined with the growth of sophisticated cyber-breaching security programs, will probably lead to an increased incidence of identity fraud, which may threaten public confidence in long-term pension systems,” the authors wrote.

Combatting those concerns will take improved assessment and understanding of AI, along with the “growth of regulation and compliance requirements” in pension systems, according to Mercer and the CFA Institute.

Planning for the Planners

AI use will not only be consumer-facing. The technology can help workplace retirement plan advisers as well, according to Benartzi. He notes that the technology can assist in areas such as client assessment, proposals and even plan management and administration.

“The reality is that it’s very difficult to cater to small plans, and therefore a lot of small employers don’t have a plan,” he says. “But what if we can use AI to make plan advisers cater to smaller plans and smaller employers?”

Machine learning and data analytics may also help advisers identify potential clients, create proposals and even pitch their services, Benartzi says.

“Does it mean the adviser is going to disappear? No. I’m a strong believe that there is going to be a hybrid approach,” he says.

Benartzi, meanwhile, is hoping that the industry, including regulators, moves toward leveraging technology in financial planning and advisement, rather than stifling it.

“We have to remember to use [AI tools] for the right applications,” Benartzi says. “AI is not perfect. But it gives you a lot of personalization for virtually no effort, and that’s why I think it’s very exciting.”