What HR should know about using AI for benefits administration

Open enrollment can be a headache for employees and HR pros alike. Can AI help fix that?

Some companies in the total rewards and human capital management space are betting on it, investing in AI-driven technologies with the aim of helping employees make better decisions about their benefits packages. Nayya and human capital technology firm Alight are among the firms that have recently rolled out AI tools with an eye toward benefits.

Though the potential for such tools is promising, HR pros should also be aware of privacy concerns surrounding AI, particularly when dealing with employees’ health data.

Benefits use cases for AI. Alight, which primarily works with large employers on benefits administration, has been testing generative AI in a few different ways, said Eddie Pinto, VP of product management and strategy.

This past fall, the company piloted “simplified enrollment” through an interactive virtual assistant powered by Alight’s own generative AI technology. Following the introduction of ChatGPT in November 2022, Pinto’s team had been wondering whether users would be drawn to an experience “where they can still see their benefits and enroll, but we’re using a conversational interface for them to ask questions, to get decision support.”

Drawing on user data about factors such as plan options, cost changes, life events, and health claims, Alight’s AI tool determined whether an employee was currently on a plan that was the best fit for them, and was unlikely to make changes to their benefits, Pinto explained. If so, the employee would be given the option to select simplified enrollment, which would lead them to the virtual assistant. This assistant would then go over any plan changes, determine whether the employee had more cost-efficient options available to them, and recommend new benefits such as long-term care insurance, if relevant. With the guidance of the virtual assistant, the average time employees spent on open enrollment dropped by about half, from 17 minutes to eight minutes, Alight found.

Alight has also used generative AI to help craft targeted health savings account (HSA) campaigns based on employees’ claims and savings needs. If an employee recently had a baby, for example, the HSA message they receive may allude to that.

The benefits decision platform Nayya, which uses generative AI and data science, announced a partnership with MetLife in October. Employers with more than 1,000 employees who offer MetLife insurance now have access to the platform, which uses AI algorithms “to provide tailored and actionable recommendations across a full range of benefits,” according to a statement from MetLife.

A majority of MetLife employees who’ve used the Nayya platform have found it to be useful, Jamie Madden, SVP of workforce engagement and benefits connectivity, told HR Brew. She added the tool has been helpful in facilitating “ongoing engagement” with enrollees, so that they’re reminded continuously of their benefits elections, and how they may use them.

Know the risks. Pinto said he doesn’t anticipate that employers will go all-in on AI tools for open enrollment right away, given HR tends to be “risk-averse.” While Alight plans to offer simplified enrollment to all of its clients this year, he expects that one-half will decline to offer the AI-powered option.

Among the potential risks that may arise with AI in the benefits space is privacy. In 2019, before ChatGPT was even publicly available, Mark Manquen, founder of Manquen Vance, a benefit broker and consulting firm, told SHRM, “as these tools get closer to collecting, interpreting, and making recommendations based on personal claims data, there is a very real chance of sowing distrust among employees regarding privacy concerns.”

Pinto said Alight’s AI technology is walled off behind a secure server. The company is currently working on a benefits pilot with a defense contractor for the US government, and picked them as a client in part to focus on rigorous security standards: “If we can pass their security, then we can do it for anyone. So…how we’re thinking about it is we’re taking the highest bar possible.”