AI and Benefits Policy Management: Guidance on the Use of AI

Benefits of AI, such as reducing staff hours and creating an ease of locating the correct version of a policy, should be based on the facts specific to each company, though some risks may be universal, such as unintended consequences and data privacy intrusions.

Benefits programs can be one of the main elements of employee recruitment and retention. They can also make up an important portion of a company’s budget. “Yet, in many organizations, benefit programs are not subject to internal governance and oversight procedures….”[1] As corporate governance business processes become part of plan litigation, and because benefits may be an area ripe for budget trimming, management of employee benefits policies is attracting more attention.

In the recent past, we’ve written about how recordkeeping can help plan sponsors defend litigation based on a breach of the duty of prudence, one of their fiduciary duties. “That duty, owed by a fiduciary to its beneficiaries is context specific…. Since these cases turn on whether a plaintiff can sue for fees they feel are unreasonable, ensuring proper recording of the context involved in fee decisions may be key.”[2]

Many corporations use a policy library to organize, track and maintain their corporate policies. A policy library is a collection of all of the relevant policies for corporate governance and internal administration for a corporation. Sometimes these libraries are housed in an intranet, such as SharePoint, others may be physical copies. Either way, maintaining the policy libraries to reflect any amendments or changes to policies to conform to new regulations or reduce risk trends is essential. Many corporations use software and SaaS to assist in this management.  Common software products for policy maintenance include: DocTract, LogicGate, Document Locator, PowerDMS, and PolicyHub. Many of those programs are cloud-based solutions that automate the tasks of policy maintenance.

Some companies may consider their benefits policies to fall into employee-based policies (rather than corporate governance). That means those companies warehouse their policies in their human resources handbooks or an area on a shared drive devoted to human resource policies. This may create issues if details on the corporation’s fulfillment of its fiduciary responsibilities are located in multiple locations: human resource handbooks and the governance library. An integrated policy library can be an important point in a compliance plan.

However, keeping up with an integrated, enterprise-wide policy library can tax already understaffed benefits departments. There may be a solution. Some companies are turning to artificial intelligence (AI) to help them keep up with policy management, at least in terms of keeping a policy library updated. Corporations already may be using AI in their corporate governance for risk management and fraud detection.

AI can assist in policy management by indexing and integrating policies across departments and shared files. This can help sunset (or archive) previous versions of policies that have been updated or changed. Similar to how corporations use programs such as DocTrack and PolicyHub, AI programs for policy management can be run via an embedded AI plugin[3] or through software programs.

With that said, using AI for benefits policy management is not without concerns. “[T]he technology poses ethical, legal, and compliance risks when not appropriately governed.”[4] Plan sponsors should consult with their legal counsel and compliance team to assess the risks of using AI for their benefits policy management.

One suggested area of conversation with compliance counsel includes whether the risks of using AI are overcome by the benefits of using it. Because AI can mitigate risks, thinking through current and future benefits of using it is essential. Benefits of AI, such as reducing staff hours and creating an ease of locating the correct version of a policy, should be based on the facts specific to each company, though some risks may be universal, such as unintended consequences and data privacy intrusions.[5]

While many industries have begun to develop their own set of best practices for using AI, common cautions include the four T’s: transparency, testing, tools, and training.

· Ensure transparency by making it clear who and how the systems were created.

· Continuously test and revise AI systems to ensure they are accurate and valid. This testing should be documented and/or archived.

· Identify the specific tools that will be used and the intended purpose for each of the tools, to prevent overuse of the system. This can assist in protecting data privacy as well.

· Clearly document how an AI system will be trained and plan for how that training will be revised.






These articles are prepared for general purposes and are not intended to provide advice or encourage specific behavior. Before taking any action, Advisors and Plan Sponsors should consult with their compliance, finance and legal teams.

Back to Blog

Latest Entries

Need a Proposal?

Before leaping into the unknown, we recommend a thorough examination of your plan. Because we are experts in the field, we know the marketplace and know what your existing vendor is capable of offering.  Through this examination, we can help you optimize the service you receive.

get xpress proposal