By the AI for PI Community
What does it take to enter the Prevention Era? At AI Summit 2026, held October 6–7 in Atlanta, that question led to discussions about how claims are created, how providers and payers work together, and how organizations put AI into practice.
Across the sessions, speakers described both meaningful progress and persistent obstacles. AI is helping teams review records, prioritize investigations, and reduce manual work. Yet legacy systems, competing incentives, limited trust, and difficulty measuring prevention continue to shape what organizations can achieve.
Six themes emerged from those conversations, with clear implications for payment integrity and special investigation unit (SIU) leaders.
1. Prevention starts with influencing how claims are created
In his opening address, AI for PI Community founder Dr. Musheer Ahmed challenged attendees to consider how payment integrity can influence claims at creation. Earlier intervention creates opportunities to address coding, policy, and clinical context before an error becomes a denial, appeal, or recovery.
Dr. Sai Balasubramanian connected that shift to the broader economics of prevention: recovering an overpayment also involves investigation, administrative effort, and potential damage to provider trust. Preventing the problem can spare both organizations that work.
Making prevention work at claim creation requires both payers and providers to be ready. Aetna’s Jenny Sobiech described two barriers: providers may hesitate to share clinical data upfront, and some lack the technology to act on earlier feedback without adding administrative work. Pati Serio similarly noted that the appeal of pre-claim technology has not necessarily translated into provider adoption. Together, their observations highlight a practical requirement for earlier intervention: build trust with providers and make guidance easy to use within their existing workflows.
2. Provider education needs to be specific, actionable, and followed through
Dr. Priscilla J. Alfaro made a practical case for provider education as a core payment integrity capability. Physicians are trained to deliver care, and billing rules, changing policies, and payer-specific requirements can create gaps that education helps close.
Her guidance went beyond sending letters. Education should reflect the provider’s specialty, patient population, geography, and applicable contracts. It should reach the person who can change the billing behavior, clearly explain the issue, and establish a timeline for follow-up. Facilities, including outpatient settings, also deserve attention.
Speakers acknowledged that delivery matters. Sobiech described providers who welcome conversations with clinicians and coders but are less receptive to letters. In the government programs panel, Rob Bledsoe of Blue Cross Blue Shield of South Carolina described targeted prepayment review focused on particular issues, alongside education and opportunities for providers to ask questions.
Most providers want to do the right thing and focus on patient care. Effective education makes billing expectations easier to understand and act on, with opportunities to ask questions and follow-up that helps resolve issues before they become recurring denials or disputes.
3. Convincing documentation requires independent verification
During the AI-generated claim panel, speakers showed how readily available AI tools can create convincing claims and medical records that challenge existing detection methods. ” Highmark’s Kurt Spear described controlled testing in which his team created synthetic medical claims using readily available AI and successfully passed them through its adjudication system. He also described generating medical imagery that appeared credible to clinical reviewers.
These examples show that even a complete medical record with details that appear to support the claim may not be enough to verify that the services actually occurred.
At the same time, speakers distinguished legitimate AI use from improper billing. Maria Young of Zelis emphasized changes in billing patterns, frequency, and coding intensity. Tom Martin of AmeriHealth Caritas noted that higher coding levels could also reflect previously under-documented care. A change warrants examination; it does not establish misconduct.
Wendy Luc-Pepin and Jennifer Frost reinforced the importance of traditional investigative work: contacting members, verifying services and locations, and evaluating whether the clinical history supports the encounter.
As AI makes documentation more convincing, investigators need to look beyond the record itself. Combining pattern analysis with independent verification helps teams distinguish legitimate care from fabricated claims and focus investigations where the evidence warrants them.
4. AI delivers more value when teams rethink the workflow
Across sessions, speakers challenged organizations to reconsider how work should happen. Adding AI to a longstanding process can improve individual tasks, but larger gains may require changing the sequence of work, handoffs, and decision points.
Lynn Garbee of Judi Health questioned the administrative waste created when claims cycle through denials, corrections, resubmissions, and payment. Serio urged teams to examine the right intervention points. Sobiech offered a practical approach: define the ideal future process, then determine whether the existing workflow can evolve toward it or needs to be rebuilt.
Concrete examples showed where organizations are already making progress. Spear described tools that synthesize records, claims, and policies for reviewers. Frost discussed using AI to triage complaints alongside claims data. These applications help teams focus their expertise where it is needed.
The opportunity is to use AI to reduce the steps, delays, and repeated work between identifying an issue and resolving it. That requires operations and technology teams to design workflows together, with payment accuracy and better use of staff expertise guiding the changes.
5. Prevention needs a measurement approach finance can support
One question kept returning: how do organizations receive credit for an error or overpayment that never happens?
Alfaro discussed comparing cohorts and tracking changes in provider behavior. Sobiech described the ongoing work of helping finance partners recognize prevention’s value, including the operational costs avoided when payment is correct the first time.
Joanne Galimi of Gartner encouraged organizations to examine performance across the payment integrity lifecycle. Disputes, retroactive adjustments, and time to identify issues can help reveal whether an investment is improving operations.
Jenny Karuna of Katch added a useful discipline: trace each AI experiment or investment back to the business outcome it is intended to improve. Adoption and usage alone do not demonstrate value.
Building support for prevention means agreeing with finance on how success will be measured from the outset. Evidence of sustained billing improvements, fewer disputes, and reduced administrative effort can help demonstrate the value of getting payments right the first time.
6. Human expertise and active governance make progress sustainable
Governance appeared throughout the Summit as an ongoing operational responsibility. Michael J. Velezis of Blue Cross Blue Shield of Alabama described a cross-functional AI committee, risk assessments, employee education, and recurring accuracy and bias audits. Galimi emphasized decision rights and the ability to stop investments that are not delivering value.
Karuna translated human oversight into concrete design questions: can a user understand, correct, and override the AI? She also challenged organizations to preserve the experience that makes oversight meaningful. Teams need opportunities to examine raw data, develop expertise, and capture institutional knowledge.
Serio and Sobiech connected this to change management. Employees need to understand how their roles will evolve, receive relevant training, and see how new tools fit their daily work.
Sustaining progress requires people who can evaluate AI’s output, question its conclusions, and act when something is wrong. Organizations need to keep developing that expertise as they automate, with clear accountability and ongoing review built into everyday work.
Bringing the conversation back to work
The Summit’s discussions made the Prevention Era more concrete: intervene earlier, communicate clearly with providers, verify evidence, and measure whether changes improve payment accuracy and reduce unnecessary work.
For leaders returning to their organizations, a useful starting point is one recurring problem, one accountable team, and an agreed way to assess progress. The AI for PI Community will continue these conversations as members put ideas into practice, share results, and work through the questions that remain.