A recent study by CLARA Analytics has identified new potential in the use of artificial intelligence for the early detection of fraud in property and casualty insurance claims. The research suggests that machine learning models can flag suspicious activity as early as two weeks after a claim is filed, significantly faster than traditional fraud detection methods.
Research Overview
The study, finalized in November 2024, analyzed 2,867 insurance claims filed between 2020 and 2024. CLARA Analytics used an unsupervised machine learning approach, specifically cohort modeling across claim development periods. This technique allowed the researchers to detect patterns in cost, treatment behavior, and relationships among claim participants that may signal fraud.
Key methodologies included:
- Identification of outliers in cost and treatment.
- Mapping of network connections between attorneys and medical providers.
- Use of unsupervised learning to detect potential fraud indicators without relying on predefined markers.
Key Findings
Several significant outcomes emerged from the study:
- Early detection: The model identified suspicious claims as soon as two weeks after the initial filing, ahead of standard investigative timelines.
- SIU referral correlation: Nine percent of open claims were flagged as having high potential for referral to Special Investigation Units (SIUs), with strong alignment between model predictions and actual adjuster referrals.
- Geographic insights: Michigan and Arizona were noted as having the highest concentration of claims with potential fraud indicators.
- Network analysis: The model’s network analysis revealed connections between attorneys and providers, which are often overlooked using traditional methods.
Broader Industry Context
The Federal Bureau of Investigation estimates insurance fraud costs the U.S. property and casualty insurance industry approximately $40 billion annually, not including medical insurance. These losses may contribute to higher premiums for consumers.
The study also referenced the “Sentinel Effect,” a behavioral concept where individuals alter their actions when they know they are being observed. In the insurance context, this could mean that companies known for proactive fraud detection may deter fraudulent activity before it begins.
Technology and Future Applications
CLARA Analytics is continuing to refine its fraud detection capabilities by expanding its network analysis to include both medical and legal data sources. The company states that its AI-driven system, supported by agentic reasoning, is designed to interpret complex claims and generate actionable insights for insurers.
According to the study, the use of unsupervised learning enables the identification of novel fraud patterns that do not rely on historical indicators, potentially offering a more adaptive and comprehensive detection method.
The CLARA Analytics study presents a data-driven perspective on how artificial intelligence can support early fraud detection in insurance claims. Through machine learning and network analysis, the research suggests a path toward more timely and accurate identification of suspicious activity, potentially reducing the financial impact of fraud on the industry.
About CLARA Analytics
CLARA Analytics is the leading AI as a service (AIaaS) provider that improves casualty claims outcomes for insurance carriers, MGAs, reinsurers, and self-insured organizations. The company’s platform, CLARAty.ai, applies image recognition, natural language processing, and other AI-based techniques to unlock insights from medical notes, legal demand packages, bills, and other documents surrounding a claim. CLARA’s predictive insights give claim professionals augmented intelligence that helps them reduce claim costs and optimize outcomes for the carrier, customer and claimant. CLARA’s customers include companies from the top 25 global insurance carriers to large third-party administrators and self-insured organizations. Founded in 2017, CLARA Analytics is headquartered in California’s Silicon Valley. For more information, visit www.claraanalytics.com, and follow the company on LinkedIn and @CLARAAnalytics.
Get the latest insurance market updates and discover exclusive program opportunities at ProgramBusiness.com.