Six state agencies have chosen AI for Medicaid eligibility work, Stanford researchers say

Six state agencies have decided to use AI tools in Medicaid eligibility work, and 21 states are considering them, according to research summarised by Stanford HAI. With new federal requirements bringing more frequent redeterminations, the choice is not simply whether to automate, but where AI can help without compounding harm.

Stanford HAI Watch analysis

What happened

Stanford HAI’s page describes research by Michelle Mello and Himaja Nagireddy on AI’s role in implementing Medicaid work requirements. It says 44 state Medicaid programmes must redetermine millions of beneficiaries’ eligibility by January 2027 under criteria in the 2025 budget reconciliation law, known as HR 1. Among the changes are work requirements and redeterminations at least twice a year, rather than annually.

The page says six agencies have decided to use AI tools so far, while 21 states are considering them. It does not identify the tools, explain what tasks they will perform, or report results from deployments. The researchers warn that overly broad or hasty use could compound the harms of HR 1.

Why it matters

Eligibility reviews are already an administrative burden. AI might help agencies manage a heavier workload, but decisions about benefits affect people’s access to healthcare. The same technology could also make a difficult process harder to challenge if its role is unclear or its decisions are poorly handled.

The figures describe agency decisions and consideration, not successful deployments or proven savings. That distinction matters: “using AI” can cover very different jobs, from paperwork support to influencing eligibility decisions, with very different consequences.

Our read

The key question is not whether states adopt AI, but what they let it do and how people can contest mistakes. Stanford HAI’s summary raises that question, but the available details do not tell us which systems states are choosing or how they will be governed. The paperwork may be growing; accountability should not be filed under “later”.

What to watch

• Which tasks the six agencies plan to use AI for, and whether it informs or determines eligibility.
• What safeguards and appeal routes states provide for people affected by automated processes.
• Whether the 21 states considering AI proceed, and what evidence they publish about results and harms.

Discussion spark: Should states use AI to help manage Medicaid eligibility reviews if a human remains responsible for the final decision, or should these decisions stay entirely out of automated systems?

Independent WittyWires tracker for public updates about Stanford HAI. Not affiliated with or endorsed by Stanford HAI; this is not an official account.