A recent study by CLARA Analytics shows that AI tools can flag possible fraud in property and casualty claims just two weeks after filing, much faster than traditional methods.
The research reviewed 2,867 claims from 2020 to 2024. It used unsupervised machine learning to group similar claims over time.
This method helped spot unusual costs or treatments and track links between doctors and lawyers that could suggest fraud.
The model flagged 9% of open claims as likely cases for Special Investigation Unit (SIU) review. Michigan and Arizona had the highest rates of fraud indicators.
The model’s early warnings closely matched actual SIU referrals made later by adjusters. In some cases, the model caught potential fraud within two weeks of the first report.
CLARA’s team used cohort modeling and network mapping to show how doctors and lawyers might be connected across claims.
This gave a clearer view than traditional fraud checks. Their network analysis is expanding to include more medical and legal records to uncover hidden links.
The FBI reports that fraud in the insurance industry costs about $40 bn per year, not counting medical insurance. These losses often lead to higher costs for policyholders.
The study also showed that good fraud systems can discourage dishonest claims. If people know their claims will be checked, they are less likely to try fraud. This “Sentinel Effect” helps reduce fraud across the board.
What makes this approach useful is that it doesn’t depend on old fraud signals. It can find new patterns that don’t follow past examples.
Pragatee Dhakal, Director of Claims Solutions at CLARA Analytics
CLARA’s AI system uses this data to help insurers make faster and better decisions during claims. Experts believe this approach can change how the industry fights fraud by combining strong data tools with human judgment.
