Breega Watch #5— Your Monthly Insider’s Guide to Tech Verticals. Why is this sector grabbing our attention ? Why now? What investment opportunities do we see arising ? This month, we’re exploring Insurance Fraud Detection. Let’s dive in.
Crafted with 💚 & 🧠 by Benjamin Deplus & Paul Mussillon.
Why Insurance Fraud matters?
Insurance fraud isn’t just an inconvenience; it’s an expensive problem. And ultimately, the losers are you and me. Honest policyholders, paying the bill.
The global cost of fraud in car, home, and business insurance reaches a staggering $45 billion. In Europe, about 10% of all claims are fraudulent, and that increases by 15% year on year — with Spain alone handling around 60,000 fake claims per year. It’s a costly game. 15% of your car insurance premium actually goes towards covering losses due to fraud.
But as fraudsters get more creative, more powerful too become the technologies fighting back.
How they game the system?
Some of the most common car insurance scams include:
- Users: Not including a driver of the same household in the policy.
- Usage: Underreporting the mileage to reduce premiums.
- History: Conveniently not mentioning past accidents.
- Location: Claiming the car is parked in a safer (cheaper) location.
Fraud happens at every stage of the policyholder journey : starting right from subscription.
Existing players are already addressing fraud detection at every step of the customer journey with claim management being the most crowded segment.
But at Breega, we see a second wave of innovation coming, sparked by the usual suspect… GenAI.
Welcome genAI, for insurers to see through the haze
GenAI is a clear opportunity to improve fraud detection accuracy and adoption is exploding.
In 2024, 83% of anti-fraud professionals predict they’ll be using AI-driven tools by 2026 (compared to just 18% in 2023). GenAI could not only boost fraud detection accuracy by 40%, but also help reduce false positives, cutting those cases by 45%.
But interestingly, while AI is proving to be a powerful ally, insurers aren’t always eager to point the finger at fraudsters : in fact, 37% of insurance professionals prefer not to allege fraud in some cases.
“But why?”
Insurance fraud management is driven by contextualization: historical relationship with the customer and its total value (eg. other financial products subscribed) need to be factored in. Often, the cost of losing a valuable customer outweighs the risk of letting a fraudulent claim slip through.
Industry Specific LLMs: The Big Opportunity
At Breega, we’re bullish on the potential of industry-specific LLMs for insurance. Training insurance-specific language models (LLMs) with industry-specific data is crucial to achieve better fraud detection and decision-making.
We’re expecting to see two types of companies emerge, with no player identified in Europe yet!
Infrastructure : With insurance-specific LLMs trained with real insurance data, allowing to understand the unique context of claims and policies. In short, we’re talking about building the Hippocratic of the insurance sector. We expect these companies to raise very large seed rounds (e.g. Mistral in France).
Applications : We can anticipate that the first use cases of fraud detection embracing these industry-specific LLMs will be predicting risk and adjusting pricing at subscription, as well as performing contextual analysis when processing claims, optimizing fraud detection efficiency.
But here’s the challenge: to train AI, data needs to be collected and fed but insurers are understandably wary of sharing sensitive information about their customers. It will therefore be key for startups to answer the data-related fear of insurers, and some are already doing a good job at it!
In short, we’re really excited about up and coming AI-driven insurance fraud detection solutions so if you’re building a company in the space, reach out to benjamin.deplus@breega.com or send us your deck here!
