"Closed" AI systems have made generative AI practical for use with benefits administration without the risk of exposing claims data or other personally identifiable information to public access. A closed system is a must-have if you plan to use generative AI to support benefits selection.
Getting your Trinity Audio player ready...

Author: Rebecca Starr

null

This article continues Gallagher's series on artificial intelligence (AI) in human resources. Here we examine the use of AI in benefits administration, associated benefits and a few notes of caution.

Previously, we addressed the human aspect of HR, the fundamental risks of generative AI and best practices, how experimentation with AI can unlock real value for HR and the pros and cons of open, closed and hybrid models.

AI applications for benefits administration

Consider these possible uses of generative AI with third-party benefits administration solutions that go beyond chatbots and decision-making tools.

  • Automated benefits enrollment. Guide employees through enrollment and provide personalized recommendations based on available data and real-time support.
  • Personalized benefits recommendations. Analyze employee data (age, health, family makeup, past usage) to suggest best-fit plans for an individual.
  • Claims processing and management. Automate non-carrier claims processing workflow, such as tuition reimbursement programs, by verifying information and flagging irregularities or issues for review.
  • Benefits communication and education. Engage and educate employees about their benefits with personalized information to improve employee understanding and appreciation of their benefits.

    • Compliance AI Solutions

    AI that ensures real-time HR compliance is a dream come true for HR leaders; however, closed AI models can't know about new laws without an informed human updating and interpreting the laws for the training dataset. Open AI systems might more easily identify changes but risk misinformation. Of the 150 generative AI models in the market today, four represent the bulk of the market. All four face lawsuits associated with compliance, hinting at the complexity of this function. We can envision an AI-based tool that monitors and updates policies — and some decent products are available. However, such tools likely carry hefty price tags that could put them out of reach for most small- to mid-market organizations.

  • Data analytics and reporting. Provide insights based on aggregated benefits use to help HR optimize offerings and identify cost-saving opportunities.
  • Employee wellbeing programs. Help design and manage health and financial wellbeing programs by tracking employee health data and suggesting relevant wellbeing activities, resources and interventions. Let AI aggregate data and potentially recommend whether an earned wage access program might benefit your employees.
  • Regulatory compliance. Ensure benefit programs comply with privacy and data protection laws (assumes the program is constantly fed updated laws and policies (see Compliance AI Solutions to the right).
  • Feedback and continuous improvement. Collect and analyze employee feedback on benefit offerings. Use sentiment analysis, when permitted, to gauge employee satisfaction and identify areas of concern.
  • Integration with HR functions. Integrate AI with other HR solutions to boost the effectiveness of those functions. Some service providers use AI to personalize their call center support to the client's culture and to identify and strip out protected health information for employer manager review and evaluation.

Advantages of generative AI for benefits administration

Using generative AI in a closed environment to support benefits administration can increase efficiency, reduce costs, ensure compliance and improve employee satisfaction with their benefits. Here are a few examples we've seen:

  • Right-sized coverage. A first-time HR tech buyer uses AI to analyze employee family size, salary and health claims history to make personalized suggestions before employees make their final elections. On average, employees choose a lower-cost plan, right-sizing their coverage to avoid paying for insurance they don't need.
  • Claim fraud mitigation. A global aerospace tech company that uncovered employee fraud associated with tuition reimbursement is exploring AI-driven optical character recognition technology to flag fraud.
  • Increased employee satisfaction. A mid-market employer customized benefits enrollment communications based on employee age and number of dependents to highlight plan and voluntary benefits options. A post-enrollment survey showed increase in employee satisfaction with benefits options. The employer plans to analyze retention data to identify a possible correlation between benefits satisfaction and retention.
  • Support for wellbeing. AI analysis of claims data revealed a roughly 10% incidence of pre-diabetes among employees. In response, the organization replaced sweet snacks in breakroom vending machines with healthier options.

As more organizations experiment with AI applications for benefits administration, we expect to see more success stories. At the same time, we acknowledge the risk of using generative AI with benefits. While no client has shared bad outcomes, it's likely they exist.

Benefits-specific concerns of generative AI

Disadvantages of AI for benefits administration include the core risks we covered in a previous post: Accuracy, bias and compliance. Beyond these points, consider the following potential concerns:

  • Data quality. Is the training dataset current and accurate? While a closed system should eliminate accuracy concerns if you're confident in your data, we recommend extensive testing with a control group before rolling out the application to employees. We know of one case where the algorithm of an early-stage AI product unintentionally pushed employees to select the highest deductible plan offered. The organization quickly discovered and corrected the problem.
  • Transparency. Employees may be concerned about "Big Brother's" involvement in their benefits. Emphasize to employees the confidentiality of a closed system and that AI tech is generating recommendations — not HR staff or insurance carriers. Invite employees to talk to an HR or benefits resource person about AI-generated recommendations to keep the human element in the mix.
  • Integration and security. Engage your IT experts in discussions about integrating employee data with plan data to enable benefit recommendations and keep your IT team involved every step of the way. A closed system means your data is behind a firewall, but insufficient security can lead to a breach of sensitive employee data. Even in a best-case scenario, with a security issue you risk losing your employees' trust in the organization's ability to manage their information responsibly — possibly the costliest outcome.
  • Cost. Licensing, customization, complexity, testing, training and maintenance can impact closed AI costs. We've seen reports of cost to build custom generative AI ranging from under $10,000 to several hundred thousand dollars. As with any HR technology purchase, first analyze your needs and business goals. Consider hiring a third-party expert to guide you.

Start small and build to a holistic approach

As we've emphasized throughout our discussions on generative AI to support HR functions, it's OK to take baby steps. Pick one or two options, starting with low-risk applications such as benefits education and communications or analysis and data reporting. Build from there as you become more comfortable with generative AI and the technology improves.

Consider one or two of these benefits' best practices as a place to start to incorporate AI:

  • Consumer accounts technology (spending, tuition, employer-based such as fitness)
  • Benefits technology (enrollment, decision support, reporting and analytics, retiree support)
  • Benefits engagement (communications, total rewards, virtual benefit fairs)
  • Year-round use (ongoing decision paths, claims-based feedback, health care navigation)
  • Virtual assistants
  • Content personalized to the individual employee user

Create a framework for the future

Incorporating AI into your benefits administration function may seem scary. Yet it's worth putting your toe in the water to maintain your competitive edge. AI in HR is evolving, and the questions and decisions you face now may seem irrelevant in a few years. However, new questions and challenges will arise as the technology evolves. You'll thank yourself for creating a framework now to guide your decision-making in the future.

Document the factors that drive your decision-making and engage all key stakeholders in the conversation, including HR/benefits, IT and legal teams. You may want to involve your chief financial officer to track return on investment. If all this feels too overwhelming, take heart. A third-party consultant can guide you in creating a decision-making framework and provide insights on AI solutions in the market that best fit your needs and resources.

Contact Gallagher if our HR technology consulting team can assist you in your journey.

Suggested reading

Palamarchuk, Natalia. "How Much Does AI Cost? Pricing Factors and Implementation Types Explained," Flyaps, 19 Mar 2024.

Author Information

Disclaimer

Consulting and insurance brokerage services to be provided by Gallagher Benefit Services, Inc. and/or its affiliate Gallagher Benefit Services (Canada) Group Inc. Gallagher Benefit Services, Inc. is a licensed insurance agency that does business in California as "Gallagher Benefit Services of California Insurance Services" and in Massachusetts as "Gallagher Benefit Insurance Services." Neither Arthur J. Gallagher & Co., nor its affiliates provide accounting, legal or tax advice.