Over the last several months, you may have felt a mixture of amazement and apprehension over the hype surrounding ChatGPT, the generative artificial intelligence (AI) tool, and its potential impact on everything from homework assignments to jobs.
Experts in the employee benefits and employment arena say there is an application for tools like ChatGPT in benefits but warn employers and plan sponsors to tread carefully.
What Is ChatGPT?
ChatGPT is a chatbot that was developed by OpenAI and became available for free to the general public in November 2022, according to an article by McKinsey & Co. It is a form of generative AI that uses machine learning to formulate a response to a user’s request.
How Could ChatGPT Be Used in Employee Benefits?
We thought a good place to start would be to ask the tool itself. We input the prompt “In 250 words, explain how ChatGPT could be used in employee benefits” and received a summary that began with
ChatGPT provided the following specifics.
*Act as a 24/7 helpdesk: Employees can ask questions about their benefits and receive “prompt and accurate” responses, reducing the need for manual inquiries.
*Assist in benefits enrollment: The tool could guide employees through enrollment procedures, explain benefits options and provide personalized recommendations.
*Provide benefits-related notifications and reminders: This could include open enrollment periods or deadlines for submitting claims.
*Assist in managing benefits: The tool could provide information about how to file claims, track reimbursement status and assist with common inquiries about plan provisions like dependent coverage, coverage limits and in-network providers.
Then we asked some experts for their thoughts.
Concerns
“The issues are going to be confidentiality, number one, and then inaccuracy is number two,” said Karla Grossenbacher, an attorney at Seyfarth Shaw. Grossenbacher, who is a partner in the firm’s labor and employment practice and leads its privacy team, wrote a blog post for Bloomberg Law about risks to employers when employees use ChatGPT.
ChatGPT is publicly available, so using it for employee health information creates risks under the Health Insurance Portability and Accountability Act (HIPAA). “What you enter into ChatGPT—You’re disclosing it, you’re sharing it. It’s not confidential,” Grossenbacher said in a telephone interview.
Compliance with the Employee Retirement Income Security Act (ERISA) is an area where inaccuracies could create huge headaches for plans. The law is very specific, and compliance is extremely technical, Grossenbacher said. When a benefits case goes to litigation, the courts will look at not only the plan documents but additional written and oral communications, focusing on inaccuracies, she added. “Using ChatGPT to generate benefits-related correspondence does not strike me as a great idea.”
Potential Uses
Chris Chan, who is an advisor in the digital health industry and has spoken about technology and the future of work at past ISCEBS Employee Benefits Symposiums[KB2] [KH3] , pointed out that technology like ChatGPT could have an application in benefits functions that involve “answering the same questions over and over again.”
The automation that ChatGPT provides also could help bring down the costs of health concierge and advocacy programs that seek to engage and motivate plan participants to take better care of their health, Chan said. He added that rules and processes could be created so that participants are set on the right path based on their health plan, health risk, etc.
In the retirement plan arena, AI could potentially help participants with asset allocation for retirement accounts, taking into account their risk tolerance. It could even persuade people to make better retirement savings decisions. “I don’t know if you want AI to start asking, ‘Should you really be getting that coffee in the morning?’ but I think those types of nudges can be helpful for people to pause and make a decision and feel better about making the right choice in the moment,” Chan mused.
However, he added that AI has a hard time displaying empathy like a human. He pointed to health research showing that patients responded better to health messages received on a pager when they knew a person sent them rather than a machine.
Wait and See
Just as technology industry leaders have called for a six-month pause on training of AI, Grossenbacher suggested that employers and plan sponsors take a similar approach to tools like ChatGPT. “I’d be inclined to wait and see how everything shakes out with the next generation of things like ChatGPT and the outcome of congressional hearings on AI and the various agencies that are coming out with guidance before making any determination about how to use it in the employee benefits space.”
