Getting answers to questions about MO HealthNet, the state’s Medicaid program, could soon change as the state prepares to roll out artificial intelligence programs to help handle changes to the program. 

As part of President Donald Trump’s budget bill, dubbed the One Big Beautiful Bill Act, Medicaid enrollees are aged 19-64 required to prove they did 80 hours of work, school or caregiving a month, with exceptions for parents of children under a certain age and those who are disabled. 

The budget bill is forcing some of the largest changes to Medicaid in years, and states are racing to embrace new technologies to help them handle the overhaul. 

Missouri was one of only six states that reported in a recent survey that they plan to use AI to help handle work requirements for the state’s Medicaid program. 

It could show up in many different ways. Missouri said it plans to use AI for client interactions, to process documents, match data, provide extra staff support and increase automation of the program. 

But most Missouri’s 1.26 million Medicaid recipients already work or qualify for exemptions. A recent analysis of 2024 U.S. Census data found that 67% of working-age MO HealthNet patients are employed, while most others are disabled, caregiving or in school. 

Work requirements are set to go into effect in December. Although it’s unclear how exactly the state plans to deploy AI to help it handle the additional workload, Missouri lawmakers approved nearly $50 million in state and federal funds to update the state’s software and for other new mandates as part of the new law. 

While questions remain about how the technology will be used, experts say states face a massive challenge as they prepare for the changes, especially work requirements. 

Why is Missouri turning to AI for help with Medicaid eligibility? 

For years, running Medicaid efficiently has been difficult for Missouri. 

The state struggled with staff turnover and outdated technology, which led to longer waits for everything from time spent on the phone with a question to a persistent backlog of enrollment applications. 

One of the main problems is that many of the state’s computer programs can’t communicate with one another, which makes verifying someone’s data — like income or amount of hours worked — more time consuming. 

“Missouri has been kind of behind the curve on most technology,” said Elias Tsapelas, the director of state budget and fiscal policy at the libertarian-leaning Show-Me Institute. 

He said the federal government appears to be pushing states toward electronic verification wherever possible. 

“They’re really wanting states to try to make as many of these decisions electronically, basically with data that they already have access to,” Tsapelas said. “So this isn’t some enormous (undertaking) of states sending out tons of letters and requiring a bunch of new paperwork.” 

The Missouri Department of Social Services, which oversees Medicaid, requested $6.2 million in state and federal funds to build or contract a new program to handle all of the information. 

Under the federal budget bill, Medicaid patients must also go through the redetermination process twice a year, as opposed to once. Ideally, Missouri can cross-reference SNAP records, income databases and employment information with the goal of fewer people needing to submit paperwork manually, said Brian Colby, the vice president of public policy at the left-leaning Missouri Budget Project. 

“We want to make the transaction as frictionless as possible,” Colby said.

DSS also requested more than 200 additional full-time workers to help handle the changes set for Missouri over the next several years. 

“Their workload essentially has been doubled,” Colby said. 

DSS emphasized that AI tools will not make decisions about enrollment or eligibility. 

“No AI tool or algorithm makes adverse eligibility determinations or coverage termination decisions,” DSS spokesperson Baylee Watts wrote to The Beacon in an email. “The system’s role is strictly limited to data aggregation, such as identifying missing documentation or summarizing uploaded files.” 

“If automated data matching confirms an individual meets the established work requirements, the system can process that positive confirmation to ensure uninterrupted continuity of care,” Watts said, adding that if someone’s status cannot be automatically verified, a state employee will handle the case. 

Watts said if AI does determine someone isn’t eligible, a state employee must review those determinations, as well. 

“A state employee will always review the complete file, evaluate any system-generated summaries, and make the final determination before any adverse action is taken,” she wrote. 

But studies on human behavior with AI models show that humans quickly default to the suggestion AI made, said Elizabeth Edwards, a senior attorney at the National Health Law Program. 

“What the research shows is that the human will eventually also start to rely on the computer, because the computer is right most of the time,” Edwards said. “They will stop looking as carefully at them. … The idea that a human in the loop is solving the problem of any type of AI monitoring is really not sufficient.” 

How might AI be used? 

Missouri listed several ways it plans to use AI. But experts hope the new technologies will bring much-needed improvements to the Medicaid system in the state. 

Budget requests show the state is already making improvements to its chatbot software, where people can get immediate answers to basic questions, including specifics about their application or case, without needing to speak with a DSS employee. Missourians will be able to report changes to their households and fill out applications or renewals without needing to get on the phone, the department says. 

“A lot of times the reason people are calling the state (the questions) are not very difficult,” Tsapelas said. “It’s things like: ‘What’s the status of my application? What do you need from me?’ Those types of things that it seems like AI could be very helpful with.” 

The state said it plans to use AI to increase back-end automation, which could look like easily pulling up policies or federal Medicaid rules to help human employees work on more complex cases, Edwards said. 

“If they have a 300-page manual, it’s hard for workers to find what the policy is,” Edwards said. “If AI is trained on that universe of documents, it can still have errors, but it can be trained,  tested and monitored so that workers are finding the right policies.” 

But Edwards is also concerned about the possibility for mistakes — which could mean the difference between someone keeping their Medicaid coverage or not. 

“What we have seen from a lot of AI products is — at least some of the ones we have looked at — not all of them fully understand the requirements of Medicaid,” Edwards said. “The big thing really consistently is that we are always concerned about how it is being monitored and tested.” 

States expressed that concern in a survey conducted by KFF earlier this year. 

Many states shared concerns about the short timelines and inability to test new software before work requirements begin. One state said that the solutions presented by vendors planning on using AI would likely not be able to verify eligibility for people like gig workers, medically frail enrollees and enrollees who may be homeless. 

Missouri already has a partnership with an AI-powered program called SteadyIQ, which allows enrollees to link their bank accounts and gig work programs, like ride sharing, to verify their income. 

Outside of SteadyIQ, DSS told The Beacon that Missouri is finalizing contracts with new vendors with plans to integrate them into the state’s existing eligibility systems. 

States turn to tech as they race the clock 

Many states reported that they had to begin making decisions about how to handle work requirements, even without additional guidance from the federal government, due to the tight timeline set out in the budget bill by Congress. 

Some critics point to the rollout of the Affordable Care Act marketplace and the difficulty states had to meet the requirements, despite having more than three years to prepare. Officials in states like Oregon warned at the time that their software was not ready for the marketplace rollout, and residents initially could not successfully enroll online. 

“You can look at the history and see that the timeline really matters,” Edwards said. “And you can still have problems, even with longer timelines.” 

Tracking work requirements also has proved difficult for states to master. In Arkansas, more than 18,000 adults lost Medicaid coverage during the state’s brief period when work requirements were in effect before a court halted the program. 

Researchers later found the policy did little to increase employment, and a study found that coverage losses disproportionately impacted Arkansans without home internet access. In 2025, Missouri was ranked 30th nationwide for residents’ broadband access. 

Although AI programs have the potential to drastically improve the user experience for both enrollees and state employees, the cost of new technologies can quickly add up, Tsapelas warned. 

“I think AI has the potential to make things better for enrollees,” Tsapelas said. “As with all things, we have to keep an eye on it, because when Medicaid makes mistakes it is very expensive.” 

For now, Missouri remains one of only a handful of states plotting out how to use AI to handle the changes that come with the new federal law. 

Exactly how the tools will be used, and whether they make the process easier, remains to be seen. 

“If they improve the data matching, that would decrease the workload significantly,” Edwards said. “But they have to build all of that, they have to test it, and they have to make sure it works.”