Washington, D.C. — The Centers for Medicare and Medicaid Services is deploying artificial intelligence to root out healthcare fraud before bad actors ever collect a single dollar — and early results suggest the approach is working at a level that’s turning heads inside the agency.
CMS Chief Operating Officer Kim Brandt, speaking at a recent government technology conference, described a “Netflix-style algorithm” the agency is now using to screen new provider applicants. The system compares incoming enrollment data against the full history of everyone CMS has ever taken action against — flagging high-risk applicants before they can begin billing Medicare, Medicaid, or any of the other federal programs CMS runs.
The results in hospice, one of CMS’s targeted areas, were dramatic. Within the first year of using the tool, the agency removed 90% of providers it had flagged as high risk — either kicking them out when they attempted fraud or denying them enrollment entirely.
“If we can stop them from coming in, or at least identify that they’re high risk on the front end, it’s a greater chance that we can then either kick them out more quickly or keep them from defrauding us,” Brandt said.
Real-Time Claims Surveillance
The AI push doesn’t stop at enrollment. CMS is also running what Brandt called a “war room” that monitors claims as they come in — rather than waiting until after payment to investigate suspicious billing. According to Brandt, that real-time approach has saved $2 billion since March of last year. The agency says it nearly doubled its total program integrity savings last year because of these new tools.
The initiative has real urgency behind it. Multiple CMS programs saw their improper payment rates climb in fiscal year 2025. Medicaid’s estimated improper payment rate rose to 6.12%, or $37.39 billion — up from $31.10 billion the year before. Medicare Part C came in at 6.09%, or $23.67 billion, while Part D hit $4.23 billion. Though much of what gets labeled an “improper payment” stems from documentation errors rather than outright fraud, the scale has pushed CMS to act more aggressively.
A Public Comment Process Is Now Open
On the regulatory side, CMS published a request for information on February 27 seeking public feedback on how to make anti-fraud enforcement faster and more transparent. Among the questions CMS is asking: whether existing statutory authority is being used to its fullest, how to improve detection speed, and how the agency can better communicate its enforcement actions to the public. Comments are due by March 30.
For nursing home operators and other long-term care providers, the broader enforcement push has implications that go beyond fraud prevention. CMS has been tightening its oversight on multiple fronts at once — affiliate relationships, PDPM audits, and now AI-driven billing analysis. Industry watchers note that the same tools being aimed at hospice and home health today can just as easily be trained on skilled nursing data tomorrow.
That risk is not hypothetical. As CMS’s focus on affiliate risk has deepened compliance concerns for skilled nursing, facilities are learning that enforcement exposure doesn’t always start with a site visit — sometimes it starts with an algorithm flagging a billing pattern no human auditor would have caught.
The agency says its third AI tool, an internal contracts analysis system called CLAW, is already saving money on the administrative side by comparing new contracts against historical norms to detect overpricing and flag irregular terms.
Whether it’s the fraud detection engine, the real-time claims war room, or the broader RFI process, the message from CMS is consistent: the era of retrospective enforcement is giving way to something faster, more automated, and harder to outmaneuver.
