There’s a bustling information highway inside the New Mexico Health Care Authority.
The state agency administers critical federal benefit programs — including the Medicaid and Supplemental Nutrition Assistance Program, better known as SNAP — in addition to enforcing child support obligations, licensing health facilities and managing the state’s waiver program for people with developmental disabilities.
All of that translates to an enormous amount of data moving through the agency’s systems every day, Chief Information Officer Paula Morgan said.
And it’s about to grow even bigger.
Starting in January 2027, Medicaid customers ages 19 to 64 will have to complete 80 hours of work, school or volunteering per month and regularly report those activities to the agency. New Mexico is one of a handful of states already planning to use artificial intelligence to help manage the additional workload created by the new requirements, according to a survey of state Medicaid officials conducted by health policy organization KFF.
Morgan noted the state will use AI tools and automation upgrades to handle some components of the work, but human caseworkers will make final decisions.
“We want to provide the easiest transition as possible in all of this, and we want to ensure that New Mexicans realize the least disruption as possible,” she said. “[That’s] complicated, so how can we utilize technology to help alleviate the disruption?”
The U.S. Centers for Medicare and Medicaid Services on June 1 released an interim final rule to establish the standards states must use to implement the new Medicaid work and community engagement requirement, which will go into effect Jan. 1, 2027.
The new requirements, which the Health Care Authority has estimated will likely result in the disenrollment of about 89,000 Medicaid customers across the state, stems from the 2025 federal budget reconciliation bill and is intended to curb potential fraud, waste and abuse.
There are a few groups carved out as exempt from the work and community engagement requirements, including pregnant and postpartum women, disabled veterans, Native Americans and caregivers of children under 14, among others.
“This rule helps Americans build skills and independence through work, education, job training, or community service, creating new opportunities for themselves and their families,” Centers for Medicare and Medicaid Services Administrator Dr. Mehmet Oz said in a news release announcing the interim final rule.
The federal government has offered $200 million — equivalent to $4 million per state if divided equally — in grants to modernize state systems and expand administrative capacity.
The New Mexico Health Care Authority plans to share more information about the state’s plan to put the work and community engagement requirements into practice after officials have a chance to digest the nearly 400-page rule, spokesperson Tim Fowler said.
In terms of information technology, however, the state agency is already part of the way there, Morgan added.
Thanks to a previous mandate from the Centers for Medicare and Medicaid Services, the Health Care Authority already has some of the essential infrastructure to adapt to the work requirements.
“Are there gaps? Yes, there are gaps, and so that’s a lot of the discoveries that we’ve been going through,” Morgan said. “Let’s identify what we have and what we don’t have.”
The agency in March introduced an AI-powered document processing system, which can scan and identify documents like timesheets, college enrollment documents and forms verifying volunteer hours.
The state agency’s agent assist system also automatically generates comments to assist in processing cases, a standardized, AI-based system of capturing key details that aligns with new documentation standards, Morgan said.
Of course, these technological solutions come at a cost. Morgan estimated the cost of technological tools to implement Medicaid work requirements at around $10 million total.
She emphasized humans will make the important decisions, comparing the use of AI tools to movie recommendation technologies built into most streaming services — they don’t start the movie without the movie-viewer’s confirmation. The AI tools might generate comments or review documents, Morgan said, but they won’t make eligibility determinations.
“It comes down to the human, to the caseworker … because it’s the human that makes the decisions, not the AI,” Morgan said.
