Building Plan Language Using Artificial Intelligence: Man vs. Machine

By: Kate MacDonald

These days, artificial intelligence, or AI, has become increasingly incorporated into nearly all areas of our lives. You cannot turn a corner without encountering it, from the smartwatch strapped on your wrist relaying sleep metrics to your phone to the GPS in your call analyzing traffic patterns and suggesting alternate routes to work. Even sitting down on the couch to watch the news is not quite as innocuous as it used to be as stories of residents being upset about datacenters going up in their communities abound.

Meanwhile, employees nationwide have become increasingly worried that this technology will make their work redundant, but many industries have found just the opposite to be true. Individuals are developing new skills and freeing up their own resources to divert their time and efforts elsewhere, while the machines work in the background. In the health benefits industry, AI can save time and effort in developing plan language, ensure that plans are complying with current and ever-changing regulations, and guarantee that employers are abiding by increasingly complicated mental health and substance use disorder parity legislation.

However, experts must weigh the potential benefits and concerns. Let’s examine the situation.

Legislative Concerns

As with any emerging technology, there are often guardrails put in place to ensure that things do not get out of control. Notably, several states are beginning to draft legislation that does so. In 2025, Illinois became the first state to regulate and restrict the use of AI in mental health therapy via the Wellness and Oversight for Psychological Resources Act. Nevada, California, Maryland, and many other states have since followed suit.

Other states have gone further, prohibiting AI’s involvement in administrative tasks, which can have implications in health plan review and claims processing. For instance, California has laws prohibiting the technology’s use in medical necessity determinations during utilization review, while in Alabama, a provider must make the final decision when coverage is denied and not rely on AI (there must be human oversight on the technology’s usage).

There are also other concerns when implementing AI in a healthcare setting, such as confidentiality rules (chatbots are not bound by the confidentiality guidelines that therapists are, for example), the fact that AI systems inherit the biases of their training data and studies have revealed uneven treatment as a result, and AI platforms not always answering to a higher command in the manner that insurance-carrying professionals report back to licensing boards.

Resource, Not Replacement

Much of the concern in healthcare is centered around mental health. Namely, one of the biggest worries is not that jobs will be replaced by AI, but care will be. Over the past several years, people of all ages, but primarily youths, have sought attention and companionship from chatbots that can mimic human interaction; sometimes tragedy has ensued as these faux relationships cannot, at the end of the day, be real, and the scripts spiral out of control.

On the other hand, experts have realized that, with the correct controls in place, healthcare practitioners can view AI as a resource, as opposed to a replacement. For example, we have seen the following improvements stemming from the technology:

• Increased validation of diagnoses
• Enhanced access to care, especially for those without ready access to services
• New exposure to emerging technologies that were otherwise unavailable, even recently
• Resources readily available and quickly searchable
• The ability to role-play conversations at any time, which can be crucial for practice or preparation for therapy sessions

Using AI to Influence Plan Design

As AI has evolved in the realm of treatment, it only makes sense that it would also apply to the creation of plan documents. While an element of human interaction (in the form of review) is likely still necessary, once a platform is told to build out a plan within certain parameters, i.e. with the appropriate plan benefits, in conjunction with particular vendors (such as utilization managers or prescription benefit managers) AI may even reveal which vendors provide the best cost-savings to the plan and other details. Generally, with AI, it may be quicker and easier to draw up plan designs.

Once a plan has been created, AI may help employers continue operating within the parameters of parity—particularly important if an audit occurs—or if the Plan chooses to undergo non-quantitative treatment limitation testing. AI may help in the initial analysis of regulations and analysis prior to these overhauls to give plan sponsors a jump-start on these undertakings.

It may also be easier for a plan sponsor to ensure they are abiding by regulations like Employee Retirement Income Security Act of 1974 (ERISA), the Affordable Care Act (ACA), mental health parity laws, and other regulations that apply to healthcare plans. AI may also be able to show how past claims information could influence what benefits should be incorporated into a plan. A platform could also show an employer, based on past claims, what sort of stop-gap insurance should be purchased. AI can help forecast healthcare spending, and in a self-funded world, every dollar saved is critical for a plan’s assets.

It has become increasingly clear that AI is not going away. It may be in a plan sponsor’s best interest to consider whether using AI to build their plans would benefit their workflow. Jumping on the train in the early days and mastering the technology may position them for future success.