Could AI Help Bust Medicaid Scammers? Minnesota May Find Out; Government Technology, January 29, 2025

Nikki Davidson, Government Technology; Could AI Help Bust Medicaid Scammers? Minnesota May Find Out

"HOW CAN AI HELP?

The governor’s plan is to detect and flag anomalies for Medicaid providers, meaning an AI system would likely be trained to identify unusual or suspicious patterns in billing and payment data.

Suspicious patterns could include:

• Billing for an excessive number of services: Flagging providers who bill for significantly more services than their peers
• Billing for unnecessary or inappropriate services: Flagging claims for services that are not medically necessary or do not align with the patient's diagnosis
• Billing for services not rendered: Flagging claims for services that were never actually provided
• Unusual billing patterns or trends: Flagging providers whose billing practices deviate significantly from established norms or show sudden, unexplained changes

In an interview with Government Technology, Commissioner of Minnesota IT Services (MNIT) Tarek Tomes explained that this use case aligns with the state’s AI strategy of leaning into less controversial use cases that don’t reinvent any wheel, as many private-sector financial institutions already use similar technology.

“In our private lives, if we have suspicious credit card transactions, we generally get a text message asking, ‘Is this really you?’" said Tomes. “So using AI and machine learning to really look at patterns — both successful and unsuccessful patterns of transactions, and to be able to flag transactions for further review or further investigation is going to be a really important capability to add to those areas in government that have high transactions where financial benefits are paid out.”

At this point, it’s a waiting game until April or May to see if the AI pilot will be approved in the state’s budget. In the meantime, Tomes said MNIT is researching vendors and the capabilities they provide, especially in terms of low-fidelity prototypes.

If the pilot funding gets a green light from lawmakers, human beings will still play an essential role in the fraud detection process, investigating the flagged transactions for actual evidence of wrongdoing or fraud."