For more stories like this, sign up for the PLANSPONSOR NEWSDash daily newsletter.
Benefits June 2, 2025
How AI, Digital Tools Drive Potential for Decumulation Innovation
With more recordkeepers embedding artificial intelligence capabilities into their platforms, many hope it will help participants select retirement income solutions.
Reported by Remy Samuels
Art by James Yang
For retirement plan participants, making decisions like picking the best annuity option or figuring out when to start collecting Social Security can be complicated and stressful—especially without access to an adviser.
But with more digital tools available in the marketplace, and with more recordkeepers introducing artificial intelligence capabilities onto their platforms, retirement income innovation appears to be on the horizon, potentially easing the burden on both plan sponsors and participants.
Power of AI
Kevin Crain, executive director of the Institutional Retirement Income Council, believes AI can become a powerful tool in retirement services because of its ability to deliver information in a conversational way, as well as to democratize advice and retirement income planning.
Crain says while traditional planning tools may allow participants to input information and receive a projection of how much income they will need in retirement, they often lack any interaction and offer a very “static” experience for the participant. However, with AI tools, the participant is able to engage in more of a conversation, which Crain says will make participants feel more comfortable when making financial decisions.
“A lot of studies show that when you get to retirement income savings planning, retirement income projection planning [and] retirement advice, people want it in the form of a conversation,” Crain says. “They don’t just want tools handed to them. [AI] is an incredibly powerful way for providers to get to greater masses of people in a cost-efficient, but personalized way that can both inform [participants] and drive them to action.”
Another powerful aspect of AI, Crain says, is its ability to aggregate information and scenarios. While many 401(k) retirement income planning tools exist, Crain says those tools often leave it to the individual to aggregate all of their accounts and savings to create a complete picture of their assets and income needs in retirement.
Crain says about 43% of people take Social Security early, but many would likely benefit from delaying it to receive a higher monthly payment. Crain says it is likely that many people are not thinking about their various pools of assets in tandem to make informed decisions about claiming Social Security and other challenges.
“AI can do a really nimble job of consolidating different sources of information to give a much more robust retirement projection and then [recommend] a better action,” Crain says.
However, he says the power of AI is also dependent on how much a participant engages and is willing to input specific information. As with other projection tools, the more personalized information participants provide, the more specific and accurate the information can be.
Embedding AI into the Recordkeeping Platform
As a result, it is important that participants feel comfortable enough to provide more information to the AI tool and are not concerned about issues like cybersecurity. Crain argues that as long as AI is fully integrated into a recordkeeper’s system and is protected by a firewall and other data security measures, the tool would be able to automatically aggregate significant information about a participant and not require much additional input.
“I think the ideal [situation] of this, and how it will really take off, is … financial services firms writing their own versions of AI,” Crain says. “They take ChatGPT, [for example,] but they embed it inside their firewall. It allows an ability to pull in data, and the recordkeeper or financial services firm feels safe that they’re not going to have a huge breach outside of their firewall.”
Kyle Caffey, vice president of Empower’s technology innovation lab, says Empower views AI technology as central to the way it serves its customers—both in the accumulation and decumulation stages of retirement savings.
“We see AI playing a role and essentially creating a hyper-personalized journey for investors, either curating the right tools at the right time [or] the right insights based on their particular situation,” Caffey says. “At the end of the day, whether they’re interacting through one of our digital channels or face-to-face or over the phone, we want to deliver an experience that’s bespoke to their needs. … So we’re embracing AI, but we’re also doing it responsibly and mindfully.”
Caffey says Empower is focused on putting strong governance in place for AI solutions, as well as setting “clear ethical guidelines and operationalizing those principals across [its] GenAI development.”
Empower is deploying a Responsible AI platform, which includes real-time and “batch-based” safety guardrails, using tools that monitor for issues like hallucinations (inaccurate responses) or toxicity (inappropriate responses).
“We’ve also implemented a rules-based orchestrator that helps route requests safely, as well as auditing, logging and thresholds to ensure transparency and accountability throughout the AI life cycle,” he says.
Caffey adds that digital protections are integrated at every layer of Empower’s GenAI solutions—from “privacy-aware” data sourcing to ongoing risk assessments and real-time dashboards that visualize AI model performance. Empower also conducts regular testing and model evaluations and maintains human oversight as part of its operating procedures.
Caffey says while Empower serves more than 88,000 plans across a variety of industries, he finds that plan sponsors are open to implementing AI tools into their plans to evaluate their potential. He adds that plan sponsors are pleased Empower is taking a “responsible” approach to AI.
“We’re building out capabilities that are ensuring that, as we’re using or developing AI solutions, that they’re always accurate, never biased [and] they’re not hallucinating things,” Caffey says. “We’re doing that proactive mitigation as we’re embracing and moving forward in this space.”
How Plan Sponsors Feel About AI
Crain argues that the ultimate gatekeeper to AI implementation in DC plans is plan sponsors, as they need to sign off on these capabilities and allow participants access. Plan sponsors will likely feel more comfortable with AI if it sits within the recordkeeper’s protected data architecture on which the plans have already signed off.
However, if the recordkeeper can offer participants access to a tool like ChatGPT that is outside of its system, Crain says plan sponsors will likely have safety and cybersecurity concerns.
According to MetLife’s 2025 Enduring Retirement Model Study, conducted in the fall of 2024, plan sponsors believe the biggest impacts AI will have are to enable sponsors to use predictive analytics to forecast participant trends (55%), summarize large amounts of plan data and information (42%), automate complex administrative tasks (39%) and optimize decisionmaking about which benefits to offer (29%).
According to plan sponsors, AI’s biggest impact will be helping them choose investment options personalized to their participants’ specific needs (52%), enabling them to develop customized retirement strategies (49%), helping retirees choose retirement income options personalized to their specific needs (48%) and assisting employees in making other benefits distribution decisions (43%).
Overall, plan sponsor respondents expressed that they expect AI to help workers make complex decisions or realize they might need some help from a financial planner or adviser.
‘Not a Product Issue’
Roberta Rafaloff, head of MetLife’s institutional income annuities business, believes AI can especially help educate participants about how to turn their retirement savings into a steady stream of income. She says not only can AI help participants choose the right retirement income option, but it can also suggest solutions customized to their individual needs.
For example, if a participant is buying an annuity, they may want to consider whether the annuity has a survivor benefit and whether they can name a spouse as a beneficiary.
“These are questions that I’m not sure people really intuitively know they have to think about, and AI and technology can really guide people and help them think about the most critical decisions they have to make when it comes to choosing the right retirement opportunities,” Rafaloff says. “If they’re prompted with the right questions, that will be super helpful.”
Rafaloff says MetLife offers a digital retirement income tool that begins with education and brings participants through the annuity purchase process. In addition to the digital tool, MetLife has a call center to answer participant questions.
“It’s not advice, it’s pure education, but we really feel that having a broad view of retirement education is critical,” she says.
The tool explains to users what an income annuity is and gives them the ability to get quotes on and design their own income annuity by asking several questions, such as if the participant wants to protect their spouse, if they want to leave a legacy and if they are worried about their longevity.
Crain argues that AI tools fully embedded into a recordkeeping platform have the potential to drive more engagement and action from participants. For example, he says close to 60% of plans now offer a managed account, but very few participants actually use them. He argues that AI tools have the potential to drive more participants to managed accounts or retirement income solutions that a plan offers.
“With in-plan retirement income options, to me, it’s not a product issue,” Crain says. “They exist, and they’re being created left and right, … but it’s very hard for people to understand [them]. … I think AI, [with] its ability to converse at the level of the person, is a tremendous help on educating people about in-plan retirement income solutions.”
Looking ahead, Caffey says he sees AI helping participants develop more personalized financial plans, as well as speeding up the operational side of recordkeeping services. He adds that AI will make call centers more efficient and provide participants with more curated answers, so the “time to resolution” is quicker.
