Benefits Bonus: Seven Ways AI Can Be Used in Employee Benefits

[MUSIC PLAYING] Talking benefits. Benefits. Benefits. Benefits. Talking, talking, talking benefits. You're listening to Talking Benefits. The podcast brought to you by the International Foundation of Employee Benefit Plans. Every month, we dive into retirement, health care, hot topics and trends, and whatever else the benefits industry throws at us. I'm Justin Help. I'm Ann Patterson. I'm Stacey Van Alstein. Let's talk benefits. Hello, Talking Benefits listeners. This is a Benefits bonus, an extra bite sized episode to bring you some added benefits content between our regular episodes. Enjoy. [MUSIC PLAYING] Hi. My name is Kathy Bergstrom, and I'm a senior editor at the International Foundation of Employee Benefit Plans. I'd like to share my recent blog post entitled, "7 Ways A.I. Can Be Used in Employee Benefits." From crafting communication to detecting fraud, artificial intelligence, often referred to as A.I. can be a powerful tool for employers and benefit plan sponsors. Presenters and authors for the International Foundation of Employee Benefit Plans and the International Society of Certified Employee Benefits Specialists have offered several ideas at recent conferences and in publications for the use of A.I. Here's a look at some of their suggestions. One potential application is in Benefits Communication. A.I. can improve benefits communication by translating and simplifying content, adjusting the tone and format to 50 audience, and brainstorming ideas, according to Sarah Carroll, a 2025 IFC EBS employee Benefit Symposium presenter. For example, Carol, who is a senior manager of Total Rewards with Evolve, an employer-based in Denver, Colorado, said a large language model, also called an LLM, such as chat GPT, could be asked to clarify language regarding any benefit offering. The prompt could ask the tool to adjust the language to a sixth grade reading level, use plain language, and in here to the company's typical tone, Carol said, quote, "It's not only easier to read, it's empowering when people understand their options, equity improves," end quote. A.I. could further adjust the content to be used in different formats, such as emails or social media posts. Carol cautioned that A.I. generated information should be reviewed for accuracy and should never be distributed without human review. Benefits professionals who use an open, large language model also should be mindful of the data they enter into the system. Carol said in an email interview, open LLMs are more budget-friendly than a closed paid LLM, but they are more likely to produce inconsistent data and have fewer security parameters, she added. Quote, "Vendors are creating closed LLMs for HRUs specifically that are much-friendly or to the type of work we do in terms of security and compliance. Closed LLMs tend to produce more accurate results and to be more user-friendly," end quote. A.I. can also be an effective tool for benefits enrollment. A.I. is not new, and neither is the application of A.I. to employee benefits or human resources. Benefit plans have used decision support tools to help plan participants choose their benefits based on certain factors for many years, but the technology has vastly improved. A.I. can use a combination of plan data and member input to help members choose health plans, explain Julia Pesaki, who presented on the topic of practical uses of A.I. for health plans at the 71st annual employee benefits conference in 2025 in Honolulu, Hawaii. Pesaki is chief strategy officer for smart light analytics in Plano, Texas. She explained, for example, that using an A.I. tool, a participant could input information about expected events during the year, such as marriage or the birth of a baby. The tool could incorporate data, including the participants' past health care claims, drug utilization history, financial information and geography to suggest the best health plan for them. Pesaki mentioned, however, that plan members could have concerns about their data being shared with an A.I. tool. Another question is whether the tool would suggest the right plan for the member or for the plan sponsor. Another potential use of A.I. is to personalize benefits. Quote, "Plan members might already be receiving information about different programs that you have to offer them," Pesaki said. Quote, "The more information they get that is not specific to them, the more unlikely they are to pay attention to it," end quote. A.I. can help benefit plans become more targeted in meeting a plan members needs and in identifying how different benefits, such as disease or lifestyle management programs, could be useful to them. Plan administration is another area where A.I. can be applied. Plan sponsors can use A.I. to gain better oversight of their plans, Pesaki added. This could include writing service agreements with vendors such as third-party administrators and pharmacy benefit managers. A.I. can also help plan sponsors validate contract performance guarantees, ensure accuracy of claims payments, validate reasonableness of fees charged to the plan and understand plan expenditures. Another area where experts say that A.I. has an impact is data analytics. By speeding up the process, A.I. has vastly improved data analytics, which is the process of inspecting, cleaning, transforming, interpreting, and modeling data to discover trends, patterns, and other information. Pesaki said that by using data analytics to analyze healthcare claims, a plan might determine that a member is likely to become a high-cost claim in the future. The information could be sent to vendors to help determine how they might be able to intervene and help the member. Leave administration is a complicated task for employers. An expert say that A.I. can help. In an article in the July August issue of Benefits Magazine, author is Melanie Payton and Chrissy Theas explained that A.I. could assist with leave management tasks such as taking a request, tracking leaves, reviewing medical documentation, communicating updates to the employee and their manager, and ensuring that pay is appropriate. Payton and Theas, who are consultants at Brown and Brown, wrote that A.I. can analyze medical records, claim history, and workforce trends, and identify red flags, such as missing or conflicting information. For example, a claims manager could use A.I. to read medical records to find the relevant information from a doctor's note, and extract and summarize the critical details. Chat bots deployed by either the employer or a third-party vendor could help communicate with employees. Well, executed chatpots understand natural language, allowing employees to ask questions, and receive answers based on plan documents. In the future, the authors predicted that A.I. could detect early signs of burnout, such as frequent absences, erratic log-in times, or stress signals, and trigger wellness outreach, and suggest appropriate leave options or support resources. It could also help with compliance monitoring to ensure that employers stay in compliance with evolving labor and employment laws. A.I. also can be an important tool in preventing fraud, waste, and abuse. A.I.'s ability to sift through large amounts of data quickly is an important tool for detecting fraud, waste, and abuse, Pesaki remarked. A.I. can scan healthcare claims to find billing problems, such as double billing, claims paid for members who are no longer covered under the plan, or fraud schemes. Quote, it might seem elementary that your claims are being processed correctly, but claims processing is extremely complex, end quote. In an article in the July issue of Benefits Magazine, author Lerent Leor described how an A.I. driven biometric authentication system also can be used to protect plan member data by detecting issues such as forged IDs or false documents. The Internal Revenue Service is already using such a system to secure taxpayer accounts, according to Leor, who is chief executive officer of the Vika Health. The article states that taxpayers accessing I.R.S. systems are asked to verify their identities by uploading a government issued ID and taking a selfie, which the system checks using A.I. powered liveness detection to ensure authenticity. Leor wrote the benefit plans can use a similar process for their members to sign up and log into their benefit apps, which is more secure than the typical user name and password approach. Experts caution that oversight of A.I. is needed. While the applications may be widespread, authors and speakers warned employers and plan sponsors to exercise caution when using A.I. and to be mindful of their fiduciary duties to plan participants. For example, employers and plan sponsors should develop and implement A.I. use its policies with clear guidelines for how A.I. can and cannot be used in the workplace. In addition, A.I. generated benefits decisions should not be made without human oversight. Employees and plan participants should be informed about how A.I. is being used. Importantly, protected health information, also known as PHI, should not be entered into A.I. tools. And plan sponsors need to ensure that their vendors are using A.I. appropriately. Any data shared with vendors that are using A.I. should be de-identified. Quote, A.I. without appropriate oversight becomes a liability factor, Pusaki said. Thanks for listening. For more information about A.I. and employee benefits, check out the International Foundation's A.I. at G.P.T. Toolkit, available on the Foundation's website. If you like what you hear, please write us on A.I. tools. It helps others find the podcast. And subscribe to the show in your podcast app so that our episodes will automatically appear on your mobile device. Talking benefits is a production of the International Foundation of Employee Benefit Lands, the largest educational association for those working in the benefits industry. If you're into benefits, check out all the International Foundation has to offer at ifep.org. Our show is hosted by Anne Patterson, Stacey Van Alstein, and me, Justin Help. This episode was edited by Emily Yang. Today's program is copyrighted in 2026 by the International Foundation of Employee Benefit Class, all rights reserved. The opinions expressed in the podcast are those of the speakers and not to be used as legal counsel.