Can AI Both Heal and Harm Healthcare?

Boudreaux said Elevance is using a payment integrity process. “We are differentiating … between what we'll call responsible innovation and appropriate coding.” 

She stressed that Elevance supports the use of AI to improve documentation and patient care outcomes. However, when providers use AI tools to inflate revenue, the company is taking action to identify and address outlier cases and abnormal patterns. 

Boudreaux added that AI has benefits, such as streamlining prior authorization by aligning medical records and reducing administrative back-and-forth. But she cautioned against AI applications—such as those used in Independent Dispute Resolution (IDR)—that increase system burden without improving care. “We're going after that,” she said. 

Elevance also is working with technology vendors such as Pega to develop AI-enhanced platforms such as its Claim Intake Workflow (CIW), which integrates Pega AI to unify claims processing and provide insights for workforce management.

• Fraud detection to spot anomalous billing patterns.
• Claims automation to speed up claims with high accuracy.
• Prior authorization, which can lead to faster approvals for some patients, but also concerns about increased denials, according to the American Medical Association.
• Risk scoring and underwriting to offer personalized pricing based on behavioral or historical data.
• Customer service for chatbots, virtual assistants, and claim status updates.
• Predictive analytics for identifying high-risk individuals for early intervention programs. 

The U.S. Centers for Medicare & Medicaid Services (CMS) a website dedicated to AI in healthcare providing resources for healthcare individuals and organizations.

AI has already demonstrated its potential for efficiency and pitfalls. Positive examples include the following: 

• CMS Fraud Prevention System (FPS). Since 2011, CMS has used FPS to flag suspicious billing in near-real-time. In 2020 alone, it helped prevent or identify more than $1.8 billion in improper payments. The agency achieve almost $2 billion in savings over a 5-year period from stopping automated claims, according to a report from the Government Accountability Office (GAO).  Predictive analytics have allowed CMS to shift from a reactive “pay-and-chase” approach to a proactive one.
• Medicaid Claims Processing. Some insurers are using AI to automatically review Medicaid claims for medical necessity and policy alignment. The goal is to use AI and human-in-the-loop systems to scale Medicaid efficiency without sacrificing compliance. The AI filters simple claims for fast approval and flags questionable claims for human review. This auto-adjudicates claims and reduces delays in payments to providers while catching overpayments.

Some negative examples include the following: 

• Cigna’s PXDX Algorithm. In March 2023, ProPublica revealed that Cigna used an AI-based system to automatically deny certain claims based on mismatches between diagnoses and “pre-approved” treatments—often without a doctor reviewing the case. Many denials involved Medicare Advantage patients and happened within seconds, and patients often didn’t know why. Doctors filed appeals claiming the decisions ignored clinical complexity. The ProPublica report sparked Congressional scrutiny and an investigation by the U.S. Department of Health and Human Services (HHS).
• AI-Based Risk Score Manipulation in Medicare Advantage. Multiple insurers have faced investigations and lawsuits—including UnitedHealth and Kaiser Permanente—for using AI to scan medical records and inflate HCC (Hierarchical Condition Category) scores, which increased federal reimbursements. The HHS Office of Inspector General and the Department of Justice allege that some AI tools were used to add diagnoses without sufficient clinical justification. 

Use Responsibly

AI holds great promise in healthcare, but only when used responsibly. While it can improve outcomes, reduce fraud, and enhance efficiency, it must be governed by transparency, ethics, and accountability.