CLIENT OVERVIEW
The client is a large global P&C insurance and reinsurance carrier offering commercial, specialty, and select personal lines of business.
BUSINESS CHALLENGE
To reduce claims leakage and improve customer satisfaction, the insurer wanted to redefine its approach to claims severity assessment and fraud detection. Rather than working with a thirdparty vendor to detect fraud and severity signals after an issue
emerged, the insurer wanted to move beyond traditional rulesbased processes and combine structured claim attributes with AI-driven analysis of unstructured content.
In its previous workflows, the claims organization found that critical indicators of claim complexity litigation risk, or reserve adequacy were often locked in free-text documents or supporting claim documentation. This meant the insurer had to rely on manual review and interpretation of these documents, hindering a consistent ability to identify emerging severity or suspicious claims patterns early in the life cycle.
By operationalizing its ability to unearth insights from unstructured data, the insurer’s goal was to proactively identify claims requiring early intervention, improve reserve accuracy, prioritize claims for specialist or special investigation unit review, reduce claims leakage, and accelerate claim resolution. This required establishing a scalable analytics foundation that could ingest, process, and analyze both structured and unstructured data to quickly and reliably generate explainable severity and fraud risk scores.
SOLUTION OFFERED
ValueMomentum was already working with the insurer for policy system maintenance, so the carrier knew ValueMomentum could help jump-start its analytics capabilities. Impressed by the combination of analytics skills and deep domain expertise showcased in a series of claims-focused analytics demos, the carrier decided to collaborate with ValueMomentum on this initiative, starting with two high-impact use cases: severity propensity and fraud detection.
The project launched for a single specialty line of business in November 2025 and was delivered in June 2026. The scalable foundation will be extended for additional products and portfolios. ValueMomentum both designed and implemented the end-to-end analytics platform on a modern, cloud-native architecture using AWS tools like S3, Bedrock, and Cloudwatch as well as Databricks and large language models
(LLMs) like Claude. The solution combines traditional machine learning techniques with advanced natural language processing (NLP) and generative artificial intelligence (GenAI) to operationalize the insurer’s vast stores of structured and unstructured claims data.
ValueMomentum was able to greatly accelerate the processing of unstructured data by creating an intelligent processing pipeline. Instead of sending every document to an LLM, ValueMomentum developed a multi-stage analytics framework that applied NLP techniques, domain-specific dictionaries, and a comprehensive claims ontology to identify high-value claim narratives and extract meaningful business signals related to litigation, liability disputes, fraud, and more. Then, only the most relevant notes and documents were analyzed using an LLM to derive deeper contextual insights, summarize claims histories, identify severity and fraud indicators, and generate explainable risk signals that claims professionals can use to make faster, better claims decisions.
Streamlining the approach to manual review has reduced the cost and time devoted to processing unstructured data, freeing up expert claims professionals’ time to focus more on highest risk claim.
VALUE DELIVERED
Establishing a unified analytics framework that combines structured claims data with insights from unstructured sources like claims adjuster notes, investigation reports, legal correspondence, and more has provided the insurer with a scalable foundation for AI-driven decision-making across the entire claims organization. The
reusable claims ontology, AI models, and unstructured data processing framework can all be leveraged across lines of business to drive better decision-making.
Streamlining the approach to manual review has reduced the cost and time devoted to processing unstructured data, freeing up expert claims professionals’ time to focus more on highest risk claims. In addition, the client is better able to detect potential severity escalation earlier, proactively identify claims with indications of fraud, improve reserve adequacy, reduce claims leakage, improve claims handling efficiency, and more effectively triage claims that require specialist review.
