A recent FSI Predictions 2025 report from Deloitte Insights explores how advanced AI-powered technologies could help property and casualty (P&C) insurers combat the costly problem of insurance fraud. By leveraging multimodal tools across the claims life cycle, insurers have the potential to strengthen fraud prevention, improve detection accuracy, and reduce financial losses that ultimately affect policyholders.
Insurance fraud is the second-most costly white-collar crime in the United States, second only to tax evasion. The Coalition Against Insurance Fraud reports that 78% of U.S. consumers are concerned about fraud — likely because its costs are passed on to policyholders in the form of higher premiums. According to the FBI, this adds an estimated $400 to $700 annually to the average American family’s insurance costs.
For P&C insurers, recent inflation-driven policy rate hikes have already increased customer attrition. In this environment, relying on continual premium increases to offset fraud losses is not a sustainable path to long-term profitability or market share growth. Instead, the report suggests that insurers can transition from traditional, rules-based fraud detection to AI-powered multimodal systems designed to detect and prevent fraud more effectively.
The Potential of AI in Reducing Fraud Losses
In a recent Deloitte survey, 35% of insurance executives named fraud detection as one of the top five areas for implementing generative AI over the next 12 months. The report predicts that applying AI-driven technologies across the claims process — combined with real-time analysis from multiple data sources — could save P&C insurers between $80 billion and $160 billion by 2032.
Multimodal AI technologies process and integrate information from diverse sources such as text, images, audio, video, and sensor data. By merging these streams, they can deliver more comprehensive and accurate insights than systems that analyze only one type of data.
Why Fraud Detection Is So Difficult
An estimated 10% of P&C insurance claims are fraudulent, amounting to $122 billion in annual losses — about 40% of the total fraud-related costs across the insurance industry. Because policyholder interactions with insurers are often infrequent, ongoing oversight is limited, allowing fraudulent activities to go unnoticed.
The report identifies two main types of fraud:
- Soft fraud — inflating legitimate claims, such as overstating repair costs or exaggerating injuries, which makes up about 60% of cases.
- Hard fraud — deliberate false claims, such as staging accidents, committing arson, faking theft, or submitting the same photo to multiple insurers.
Growing Demand for Advanced Detection Tools
The COVID-19 pandemic accelerated digital transformation, creating new opportunities for both fraud and innovation in fraud prevention. The global fraud-detection technology market is projected to grow from $4 billion in 2023 to $32 billion by 2032. Meanwhile, regulators such as the National Association of Insurance Commissioners are increasing pressure on insurers to implement more advanced detection systems.
How AI Enhances Fraud Detection and Prevention
AI can help insurers quickly flag suspicious claims, allowing human investigators to focus on the most complex cases. When combined with advanced analytics—and deployed in compliance with applicable laws—AI can be especially effective in high-volume, high-complexity segments like property and auto insurance.
The report outlines several AI-driven techniques:
- Text analytics — Natural language processing to review claims forms, emails, and social media for suspicious language or inconsistencies, while adhering to anti-bias regulations such as the Colorado AI Act.
- Audio-image-video analysis — Speech recognition and sentiment analysis to assess customer calls; photo forensics to identify metadata manipulation or repeated use; causation analytics to validate injury claims; and video analysis to verify damage and detect staging.
- Geospatial analysis — Satellite imagery and 3D drone footage to confirm damage location and severity, reducing the need for on-site inspections in hazardous areas.
- IoT data — Telematics from vehicles and smart home devices such as water leak detectors or security cameras to verify incidents and identify staged events.Simulation models — Virtual scenarios to detect patterns of overbilling, unnecessary services, or organized fraud rings.
Combining AI and Human Expertise
For more than two decades, insurers have maintained special investigative units to address fraud. The report emphasizes that the most effective approach may be to combine sophisticated AI systems with skilled human oversight. This pairing could allow insurers to detect more fraudulent claims, save billions for policyholders, and meet long-term anti-fraud objectives. Success will also depend on attracting and retaining skilled talent and continuing to invest in automation.
Get the latest insurance market updates and discover exclusive program opportunities at ProgramBusiness.com.