The practical landscape of AI in payment integrity and VBP

Artificial intelligence in healthcare is maturing along a spectrum, ranging from targeted machine learning models to more ambitious projects aiming for broader cognitive capabilities. Narrow AI, which includes familiar techniques such as machine learning and natural language processing, is well established and underpins many current healthcare applications. Meanwhile, the industry continues to explore more generalized AI that could handle a wider variety of tasks traditionally reserved for human experts. While these ambitions fuel innovation, most healthcare organizations remain focused on concrete, incremental improvements to existing processes.

Deploying AI in real-world healthcare settings faces persistent challenges. Integrating new models with legacy systems, safeguarding patient data, and navigating an evolving regulatory framework require cross-functional teams and continued investment. Larger organizations have made better progress moving AI projects from proof-of-concept to production, but integration and security remain the top barriers. Data privacy and regulatory compliance are not simply hurdles, but priorities that shape how AI initiatives are designed and governed.

Here, we break down how Cotiviti, Edifecs, and our customers are applying AI to support two critical initiatives: payment integrity and value-based payment.

How Cotiviti’s payment integrity customers are using AI

In surveying our health plan customers about their use of AI, the most common applications focus on improving process efficiency, payment integrity claim selection, and customer service. When asked about their use cases, nearly half cited process efficiency as their main goal, followed by enhancing access to information and streamlining customer interactions. In some cases, Generative AI and machine learning are being used side by side, with both techniques supporting claim selection and documentation review.

Despite these advancements, adoption remains uneven. In our poll, we included a list of 10 common AI use cases. Cotiviti’s clients typically reported tackling two or three of those use cases, and no single use case has seen more than 35% of clients implementing it. In broader industry studies, the largest payers and providers reported they are moving between 20–50% of their AI proofs-of-concept into production, with integration challenges remaining a leading obstacle. Collaboration between payers and providers shows promise for future AI use cases, as both groups focus on similar process improvements, suggesting that shared AI solutions could help eliminate inefficiencies across the system.

How Cotiviti uses AI in chart selection

Cotiviti’s Clinical Chart Validation (CCV) solution audits a range of claims, including inpatient stays, readmissions, and complex outpatient scenarios. To perform this service, Cotiviti reviews claim data, selects claims with a high probability of coding errors and requests medical charts to validate overpayments. Machine learning is used to prioritize and refine chart requests for our certified auditors, reducing the number of chart requests and increasing the value of reviews.

Cotiviti’s approach to chart selection leverages machine learning to guide both prepayment and post-payment chart validation. The process is anchored in a “human in the loop” philosophy, where AI highlights areas of a claim or medical record for review, but trained auditors always make the final determinations. This ensures that automation does not override expert judgment, and that fairness and accuracy remain central.

By focusing on high-impact charts, Cotiviti decreases unnecessary audits, respects provider time, and fosters trust. The models are continuously improved through feedback, performance audits, and adaptation to new regulations.

Leveraging AI, Cotiviti has been able to significantly improve the precision of our medical chart selections, enabling a significant shift to prepayment. Shifting chart validation from retrospective (postpay) to prospective (prepay) review has proven especially valuable: prepay audits enable more accurate billing and upfront cost avoidance, capturing savings faster and with less administrative burden. For one client, Cotiviti’s AI helped enable 42% of chart reviews to move to prepay, yielding higher per-chart findings and faster value realization, while also growing the value of the overall program by 45%.

How Edifecs uses AI for provider network analysis

Edifecs, a Cotiviti business, applies AI to perform provider network analysis and evaluate value-based care readiness. The system produces a holistic provider score, benchmarking efficiency and quality across several dimensions including cost of care, adherence to protocols, care transitions, and payer relations. This score integrates both clinical and social determinants of health (SDOH), linking medical data with broader social context to identify root causes of performance gaps.

The analysis engine ingests claims, encounters, and third-party SDOH data, cleans and validates it, and then applies three modeling approaches. The fair provider grouping model clusters similar practices, enabling equitable comparisons across regions. An explainable boosting model breaks down predictions by specific features—such as regional crime rates—making results transparent and actionable. Finally, the aggregate model combines these measures to predict patient outcomes, including changes in chronic conditions and mortality rates.

Edifecs’ models are validated across millions of providers and rigorously tested for bias, with detailed explanations available for each prediction. This transparency helps both plans and providers understand the factors driving performance and supports value-based contracting. By including SDOH data, Edifecs ensures that assessments are not just clinically accurate but also socially informed, paving the way for fairer, more effective network management.

Looking to 2026 and beyond

As these use cases demonstrate, AI is already delivering transformation in both payment integrity and value-based care enablement. By combining machine learning, expert judgment, and the integration of clinical and social determinants of health, these solutions can improve healthcare value and quality, significantly reduce administrative burden, and deliver insights that can be acted upon faster. As payers and providers set their strategies for 2026 and the years ahead, the responsible, collaborative use of AI will remain essential in driving sustainable improvement. The future of healthcare lies in thoughtful, responsible implementation of AI, ensuring technology serves expertise—not the other way around.

On-demand webinar: Deploying AI effectively and responsibly in payment integrity

Watch the latest webinar in our quarterly Payment Integrity Pulse series as we dive deep into:

• The current state of payment integrity AI for payers
• A framework for responsible AI use
• Successful and upcoming use cases for payment integrity