A healthcare claim can look entirely legitimate: The diagnosis codes align with the procedures, and the medical record appears to support the services billed. But what happens when generative AI can create all of it?
During Codoxo’s recent AI for PI Summit, Kurt Spear of Highmark Blue Cross Blue Shield described an experiment in which his team used a publicly available genAI tool to create medical claims spanning an entire episode of care, from an ambulance encounter through surgery, rehabilitation, and durable medical equipment. The team successfully processed the claims through its adjudication platform during internal fraud detection testing.
Perhaps the most concerning discovery was how little healthcare expertise was required. The artificial intelligence tool generated the diagnosis and procedure codes needed to construct the claims, demonstrating how easily sophisticated billing scenarios can now be created.
As providers increasingly adopt AI for clinical documentation and coding, payers are turning to the same technology to evaluate claims and improve investigative efficiency. This growing use of AI on both sides of claims processing creates new challenges for payment integrity, along with opportunities to improve accuracy.
The implications extend across the payment integrity lifecycle, particularly as AI-generated healthcare fraud continues to evolve. Insights shared by leaders from Highmark, Optum Serve, Zelis and UnitedHealth Group point to five areas where healthcare payers can strengthen their strategies and take action today.
1. Monitor Changes in Coding Behavior Before They Become Larger Problems
AI-assisted documentation and coding tools are increasingly becoming part of routine clinical workflows. These technologies can help providers reduce administrative work, but they may also influence coding intensity, coding accuracy, and reimbursement patterns.
Maria Young of Zelis emphasized the importance of watching for shifts in billing patterns as AI-assisted documentation becomes more widely used. Changes in how providers code and bill for services can offer early indications of emerging fraud, waste, and abuse (FWA) issues, even when individual claims satisfy existing payment rules.
For payment integrity and SIU teams conducting fraud detection, the key is understanding whether those changes are consistent with the care being delivered. An increase in higher-intensity coding, for example, may warrant closer review if it differs significantly from a provider’s historical billing or the practices of relevant peer groups. Comparing these trends over time can help identify emerging billing issues that warrant closer review.
This analysis can also inform preventive action. An emerging coding issue may signal the need for a reimbursement policy clarification or a new claims edit. In other cases, targeted provider education could help address the issue before it becomes widespread.
2. Validate Claims Against the Patient’s Broader Clinical History
As AI-generated documentation becomes more convincing, medical record reviews may need to incorporate additional evidence to establish whether billed services were appropriate. Spear described how Highmark is involving medical directors in developing more clinically informed approaches to identifying suspicious or fraudulent claims. For example, when reviewing claims for diabetic testing supplies, medical directors can look for evidence of diabetes-related care in the patient’s broader medical history. This approach helps payment integrity teams evaluate claims within the context of the patient’s overall care.
Investigators also benefit from working closely with clinicians who can help interpret unusual findings. AI-powered analysis can then examine larger volumes of claims for similar patterns. Jennifer Frost of Optum Serve and Wendy Luc-Pepin of UnitedHealth Group reinforced the importance of traditional investigative methods, including contacting members to verify services and confirming that providers are operating at the locations associated with suspicious claims.
Payers should consider where these verification steps can be incorporated into existing investigative workflows, particularly for high-risk claims involving unusual clinical histories or potentially fabricated documentation.
3. Use AI to Make Investigations More Efficient and Actionable
Healthcare payer SIUs and payment integrity teams are under pressure to manage growing workloads while controlling administrative costs. AI offers opportunities to reduce the manual effort required to investigate suspicious activity. Several examples demonstrate where organizations are already finding practical applications.
Highmark has developed an AI-enabled tool that synthesizes medical records and claims information in the context of reimbursement policies to support payment appropriateness reviews. Another tool gathers publicly available provider information, allowing investigators to spend less time conducting routine background research. Frost also described using machine learning and predictive modeling to help triage complaints in a federal program integrity environment. By combining complaint information with claims data, the approach helps teams prioritize referrals that may warrant investigation.
For organizations evaluating similar applications, a useful starting point is identifying the investigative activities that consume the most time. AI may help accelerate initial complaint screening or the research investigators conduct before reviewing a case. The goal should be to give investigators faster access to the information they need to make informed decisions and focus their expertise on the cases with the greatest potential impact.
4. Build Accuracy and Human Oversight Into AI Workflows
As payers expand their use of AI, the quality of the findings matters as much as the efficiency gains. False positives can create unnecessary investigative work and increase friction with providers. Inaccurate or incomplete analysis can also cause teams to overlook important risks, including potential patient harm.
Spear explained that Highmark continually measures the performance of its AI tools and refines them based on the results. Luc-Pepin similarly emphasized the importance of having experienced professionals validate AI-generated findings and help improve the technology over time.
Additional considerations around AI in a claim-generation era include:
- Healthcare payment integrity leaders should measure AI performance based on the quality of its findings and whether those findings lead to actionable results.
- Investigator time saved is another important factor to consider.
- Organizations also need to train their teams to recognize the limitations of AI-generated analysis. Investigators should know how to verify the evidence behind AI-generated findings and when additional clinical review is necessary.
These practices can help organizations build confidence in AI-supported decisions while maintaining appropriate oversight.
5. Turn Investigative Findings Into Better Payment Integrity Controls
One of the greatest opportunities for payers is applying what they learn through investigations to prevent similar issues from recurring. Spear shared an example involving an unusual pattern of cataract-related services billed for patients receiving hospice care. The finding prompted further examination and raised questions about whether additional medical reimbursement policies might be appropriate. The example illustrates how an unexpected billing pattern can lead to improvements beyond the original investigation.
For healthcare payers, this presents an opportunity to apply investigative findings more broadly across the payment lifecycle. Insights from postpay investigations can inform prepay claims editing, strengthen reimbursement policies, and proactively identify opportunities for targeted provider education. Connecting these activities can help organizations address the underlying causes of payment errors and prevent them from becoming recurring problems.
Preparing Payment Integrity for the AI-Generated Claim
The growing use of AI in healthcare will continue to change how claims are created and how payers evaluate them. Maintaining effective payment integrity programs will require organizations to recognize emerging billing patterns and use clinical context to make more informed decisions. AI can help investigators do this work more efficiently.
Payers can begin making these improvements within existing operations. Stronger connections between investigations and prevention can help organizations protect payment accuracy as the claims environment evolves.
To learn more about rethinking provider engagement and to view the full-length AI for PI Summit session, visit the AI for PI Community YouTube page.