AI Dominates Program Integrity Headlines, HI Critical for Success
As the shift from pay-and-chase to pre-pay determination solidifies in the minds of program integrity professionals, this year’s National Association of Medicaid Program Integrity Conference in Portland, Oregon was filled with vision, strategy, and AI.
Gone are the days of experimenting with AI and discovering use cases. AI-enabled applications and LLM-based systems are being used to automate workflows and maximize productivity across the program integrity spectrum.
Revolutionary Documentation Delivery
Many Medicaid operations require hundreds, sometimes thousands, of pages of policy and compliance reviews of medical providers. Many states employ policy expert teams tasked with monitoring provider operations and documenting their compliance against regulations. These are typically small teams who must verify real medical outcomes are performed properly. This can be an arduous task requiring attention to detail and keen insight into medical policy. NAMPI attendees got a glimpse of new systems that adroitly receive observational input and policies; then, generate output indicating varying levels of compliance.
Reducing Manual Load
Clever software and LLMs are reducing the time needed to sift through claims. Companies demoed systems featuring the ability to use prompts in claims retrieval and search. This alleviates some of the load professionals accustomed to using macro-infused, mission critical Excel spreadsheets to analyze claims from MMIS and upstream analytics platforms. Solutions using this approach also cut the round-trip time from idea contemplation to claims retrieval from the analytics group.
Medical reviewers were awed by new technologies giving users the ability to chat with medical records and provider documentation. Yours truly and our TENEX team were featured on a panel sharing how new AI, LLM, and software techniques are used to reduce time in medical records and documentation review cycles. Products like TENEX create new opportunities for medical reviewers to dive deeper into records because LLMs and AI can analyze 1000s of pages of information quicker than humans physically can.
The Evolution of Human-In-Charge
Adoption of AI technology is very, very recent. Human-in-the-loop (HITL) is the current human involvement and verification standard being employed by AI consumers and deployers. More decisions are being influenced by some form of AI. Incorrect AI decisions may involve rework or reversion to manual processes. Users are finding the reactive determination of AI outputs to be costly.
Human-in-charge, also known as human-at-the-helm (HATH?), is a proactive approach utilizing AI’s strengths in an instructions-and-guardrails-throughout-the-workflow instead of the run-and-review-results approach many use today. HIC requires novel software approaches to make sure the human user is present throughout the process and verifies results and redirects system behavior when necessary.
Human Insight, Expertise, and Experience Guarantee Success
One thing was clear during NAMPI 2026. Regardless of the amount of AI and machine learning technology used in program integrity work, the biggest indicators of success are the people involved. AI/ML can quickly analyze and process data, but it still can’t convert data into information. LLMs can efficiently search, seek, and summarize information in reams of documentation, however LLMs don’t intuit or provide insights into actions or behaviors.
When it comes down to it: AI, machine learning, and the LLMs based upon these techniques are tools. Tools, when placed in the hands of masterful professionals, are effective for finding fraudsters and safeguarding public trust in the health system. As we like to say at Qlarant, AI+HI is the key to creating great outcomes.
