With Congress poised to reduce federal Medicaid funding by 12%, healthcare program integrity leaders face mounting pressure: maintain Medicare and Medicaid services oversight for patient care, control healthcare fraud, and safeguard taxpayer dollars—all while doing more with less.

Artificial intelligence (AI) is emerging as a vital tool in that mission. But implementing it isn’t just about choosing an AI tool. It’s about finding the right starting point, gaining organizational stakeholder buy-in, and ensuring technology enhances—not replaces—the deep expertise inside agencies.

At the 2025 AI for PI Summit, two sessions—one a cross-gov panel and the other a detailed case study—shed light on how AI is being operationalized across public health programs.

Start with the People, Then Layer in Technology

Leaders emphasized that AI tool implementation must begin with the people doing the work. One Medicaid Program Integrity team opted to explore lightweight, internal use cases first—like a “policy bot” designed to capture healthcare system institutional knowledge and help staff find the right guidance without asking overburdened colleagues.

In another program, leadership took a more structured approach, launching a statewide pilot with Codoxo’s FraudScope. But despite having fraud prevention technology in place, their processes mirrored each other in one key way: they involved staff early and often.

“We brought investigators, nurses, and data analysts into the pilot itself, they sat at the table, tested the tool, and gave real feedback. That was critical.” said one state Inspector General

Building Cross-Functional Teams to Act on AI Insights

One of the standout agency initiatives discussed following the AI rollout was a shift in team structure. Instead of keeping clinical review, data analytics, and fraud scheme investigations in separate silos, the agency moved to a team-based model—triads of a healthcare provider, analyst, and investigator working together on a case real-time from start to finish.

This model, adapted from the agency’s Medicaid Fraud Control Unit experience, ensured that each team had a full range of health information perspectives, so they could triage faster, collaborate better on claims data, and reduce case cycle times. It also made it easier to train staff across functions and build methodologies as the AI surfaced new healthcare fraud risks.

Balancing Innovation with Reality to Find Success

AI and predictive analytics adoption in state and federal agencies doesn’t happen in a vacuum. Panelists acknowledged the very real fiscal, procurement, and regulatory hurdles that come with implementing new technology.

• Budgeting: We found some agencies need a multi-year appropriation to support implementation and scaling, while others’ internal efforts faced a different friction—projects that saved money risked losing future budget allocation.
  
• Governance: Engaging with internal IT, data security, and contracts teams early was key. Our panelists emphasized how much time they spent explaining encryption protocols, access layers, and data transmission standards to get internal stakeholders comfortable.
  
• Policy alignment: AI can only flag vulnerabilities that the system knows is wrong. The most effective use cases linked claims data platform findings directly to documented Medicaid policy violations.

The agencies emphasized that ROI goes beyond financials—operational gains matter too:

• Time saved on triage and referrals
• Faster Medicare fraud scheme case resolution
• Better staff alignment and retention

Lessons for Agencies & Payers: Don’t Wait for Perfect Conditions

What made these implementations successful wasn’t perfect funding, tech stacks, or expertise—it was strategic prioritization and smart change management to get ahead of improper payments.

• Start small: one agency’s homegrown  tool didn’t cost millions. It focused on preserving institutional knowledge and solving a specific bottleneck.
  
• Build cross-functional teams: in another case, the team-based approach ensured shared ownership and faster execution.
  
• Measure broadly: Success isn’t just dollars recovered. Track reductions in case cycle time, improved triage rates, and internal engagement.
  
• Plan for scale: We heard that one agency is now planning to ingest pharmacy data next—targeting opioid overprescribing as a new frontier.

Where AI in Program Integrity is Headed

Looking ahead, attendees heard the importance of expanding the scope of AI-driven reviews:

• Ingesting pharmacy claims to begin identifying patterns in opioid prescribing and MAT services.
  
• Focusing on pre-pay alerts to catch high-risk behavior earlier in the claims lifecycle—not after millions have already been paid.
  
• Going beyond medical records and claims data and integrating open-source intelligence and business entity datasets, which is critical as fraudsters become more sophisticated and digitally agile.

Final Takeaway: AI is a Partnership

AI in payment integrity isn’t a one-click solution. For government agencies and commercial payers alike, it’s an iterative, collaborative process that works best when built around staff inclusion and training, practical, policy-linked use cases, strong internal champions and governance support, and metrics that track efficiency as well as recoveries. And most importantly, doing it in a way that respects the work your teams already do—to make them better at it.

Ready to transform your own payment or program integrity program? Visit https://pointzeropaymentintegrity.codoxo.com to learn more.