The Coalition for Health AI is convening a fast-tracked, priority team to determine how AI could best be used to implement Medicaid work requirements passed in the One Big Beautiful Bill Act (OBBBA) earlier this month.
The Medicaid work requirement policy will require able-bodied adults between 19 and 64 to work 80 hours per month, or engage in another qualifying activity like volunteering, to receive Medicaid. The “community engagement requirement,” as the bill calls it, is the major driver of the healthcare savings achieved by the bill, which is expected to reduce healthcare spending by $1 trillion over 10 years.
The Coalition for Health AI believes that AI can be used to determine Medicaid eligibility under the new requirements, its CEO Brian Anderson told Fierce Healthcare in an interview. “I strongly believe that AI can support the new eligibility determinations from the OBBBA,” Anderson said.
Anderson sent a form to CHAI community members – including tech vendors, state public health officials and employees of the Centers for Medicare and Medicaid Services (CMS) – to participate in the so-called “tiger team” roughly one week ago.
The phrase tiger team is often used to indicate high-priority tasks on a tight timeline, according to Forbes.
Anderson said roughly 150 organizations have responded to the form to signal their interest in participating in the group, including several state public health officials.
The tiger team will develop best practices for the use of AI in two cases, one for helping individuals complete their Medicaid applications and the other for helping decide if applicants meet the new Medicaid requirements.
“We really appreciated this was likely a space where AI could create efficiencies and help people complete their applications, but also make the appropriate adjudication determination, and trying to do that in a way that doesn't create increased burden for states,” Anderson said.
The tiger team will work on an advanced timeline and attempt to develop best practice frameworks within six months and finalize them within a year.
Typically, CHAI best practice frameworks go through a long process of community review and comment period. The frameworks for Medicaid work requirements are not likely to go through this rigorous community process because of the tight deadline.
“This is a little bit more of a nuanced area where we may not need as much rigor as one of our more formal best practices,” Anderson said. “This may be an opportunity to try to do something faster.”
The new Medicaid work requirements go into effect in about 18 months, on December 31, 2026, and the Department of Health and Human Services (HHS) must release guidance on the community engagement requirements by June 1, 2026. CHAI’s timeline aligns with the HHS guidance deadline.
States will also be eligible for $200 million grants for FY 2026, which will in part be based on the proportion of impacted individuals, according to an analysis of the bill by the National Association of State Health Policy. CMS will also receive $200 million to implement the new requirements, NASHP says.
“New Medicaid community engagement requirements, for example, will require comprehensive changes to eligibility and verification systems for the over 20 million people in the Medicaid expansion population,” NASHP says. “This will include establishing new data-sharing arrangements and infrastructure across state programs to correctly identify individuals who meet or should be exempt from requirements or who are meeting the monthly community engagement requirements.” The blog post notes that states will likely have to invest in new information technology infrastructure.
CHAI has not made any formal or informal promises to CMS to develop these best practices, Anderson said. He also said he does not know if CMS plans to use AI for determining Medicaid eligibility under the new requirements set out by the OBBBA.
Anderson recently spoke at the CMS Quality Conference in early July about the use of AI for prior authorization in a fireside chat with CMS Administrator Mehmet Oz, M.D., DOGE Acting Administrator and CMS Advisor Amy Gleason, and representatives of Epic Systems and Patients for Patient Safety US.
He said he has spoken with several CMS employees about the use of AI in administrative processes. CMS is also invited to participate in the tiger team.
“We welcome the administration to participate, obviously having public sector officials, in this case, the regulators at the CMS level, participating in an industry-led effort to try to understand how we can best use AI in this space would be enormously helpful,” Anderson said. “So we've offered them an opportunity to participate.”
Anderson said he started thinking about the use of AI for Medicaid administrative processes during the post-pandemic Medicaid redeterminations in 2023. States were prevented from disenrolling individuals in Medicaid during the COVID-19 public health emergency and engaged in redeterminations after the continuous enrollment provisions ended.
“Some application forms were around 10 to 15 pages, and we found lots of people were not accurately completing those applications … and part of the reason was just incomplete forms, other was insufficient information,” Anderson said. “A few of us at the time that were working in the private sector began appreciating that this might be a space where generative AI could really help in the completing [sic] of an application form. We were looking to partner with several states … that brought AI and other digital tools to help kind of navigate the renewal process that were actually successful.”
Anderson said there are existing AI solutions on the market that could help individuals apply or re-apply for Medicaid.
“There are existing digital solutions that could help on the front end an individual reapply or apply and identifying all the new information in the eligibility determinations section,” he said. “I would look to some of the commercial payers that support Medicaid. I think that they've been some of the ones that have been using that technology.”
Anderson said the Medicaid work requirement tiger team will likely rely on the foundations set by the CHAI workgroup on AI for prior authorization. Their framework so far says that when services requiring prior authorization are approved, AI can automatically send approval. When the services might be denied, AI flags the claim for human review.
Anderson said that AI for work requirements would likely mirror the use of AI by commercial payers for claims adjudication.
“At an adjudication level, there's been lots of media coverage that has shown that commercial payers do involve some kind of AI in the initial review of things that need to be adjudicated, but there's always a human in the loop,” Anderson said. “So I would say some of those existing workflows and digital technologies or digital solutions could be repurposed. You have a new kind of training set of data that you would need to take these models through, but it could be used to do similar kinds of adjudications.”
Medicaid provisions of the One Big Beautiful Bill Act, signed into law by President Donald Trump on July 4, are estimated to boot over 10.3 million Americans off of Medicaid, according to an estimate by KFF.
Many Medicaid applications may be flagged for review because the new policies are expected to exclude individuals from the program. Anderson said AI could still be useful in these cases of human review of applications.
“Hopefully, the kinds of solutions that are out there enable that human to get to the source data, to be able to make an ultimate determination, and so that kind of summarization of different data for an individual's application or renewal, an AI could also help with that,” he said. “So there's additional efficiencies that can still happen downstream, even once a human gets into the loop."
Anderson said one of CHAI’s guiding principles is to use AI to help individuals, and that remains the case with the tiger team.
“This is obviously a very sensitive and important space that affects millions of Americans. I think one of the guiding principles within CHAI is that AI can be used as a tool to help people … This seems like a good opportunity to see if we can help, if we can develop a framework that would inform an ecosystem to develop a set of apps that, ideally, where AI is helping people. … As it relates to the bill, I mean, it is the law and we need to work within the framework of the laws made by our officials. And in this case, we want to see how AI can help within the structure of the new law.”