Using an impact-driven approach, Foursys helped a multi-line insurance company reduce claims analysis time by 85% and build a business case with a return on investment estimated at 8 months.
After six artificial intelligence pilot projects that yielded no measurable results, the insurer needed to turn skepticism into trust—and technology into operational results.
Foursys combined opportunity mapping, a proof of concept for automated triage using large language models (LLMs), and financial analysis to demonstrate the potential of AI in the claims process. The initiative reduced processing time from five business days to less than 18 hours per claim and supported a business case approved by the board within six weeks.
The Challenge: Claims, Rework, and Six Unsuccessful Pilots
The insurance company faced a claims analysis process that took, on average, five business days. Undetected fraud and intensive manual rework were causing operational losses, while the claims team was operating at full capacity.
The situation was exacerbated by a history of six AI pilot projects conducted by other vendors, all of which yielded no measurable results. In addition to the accumulated investment with no return, these experiences created a trust barrier within the executive board.
Reducing analysis time, improving the ability to identify fraud, and demonstrating economic viability were priorities for securing new investments.
The main challenges:
- Low Operational Agility: Time-consuming analyses limited the team’s processing capacity.
- Fraud and Manual Rework: Difficulties in identifying inconsistencies increased operational effort and financial losses.
- Skepticism from Executive Management: Six previous pilot projects with no results made it difficult to gain approval for new AI initiatives.
- Pressure for Demonstrable Return: Continued investment depended on evidence of impact within a short timeframe.
Why Foursys?
The project required an approach that would connect artificial intelligence to a concrete business problem, without repeating the cycle of experiments with no return on investment.
Foursys applied the “AI with Impact” approach, prioritizing opportunities with measurable gains and validating their potential before proposing expansion.
The initiative was structured around four pillars:
- Business-Oriented Diagnosis: Mapping bottlenecks and AI opportunities in the claims process.
- Quick Wins with Real Impact: Prioritizing an application capable of demonstrating value within a controlled scope.
- Technical and Economic Validation: Development of a proof of concept (PoC) accompanied by the construction of a business case.
- Evidence-Based Scaling: Projecting expansion based on the initiative’s results and financial viability.
The Solution: Automated Claims Screening with LLMs
Foursys developed a PoC for automated triage using language models to semantically analyze claims documents.
The goal was to reduce manual effort and accelerate analysis by linking technological validation to operational and financial indicators.
The deliverables included:
- Structured Opportunity Mapping: Identification of process points with potential for improvement through AI.
- Document Analysis with LLMs: Use of GPT-4 and Claude for the semantic interpretation of claims documents.
- Processing and Integration: Structuring pipelines and services to work with data from legacy systems.
- Metrics Tracking: Dashboards for operational monitoring and fraud-related indicators.
- Business Case and Scalability Projections: Consolidation of the economic assessment, with a return on investment calculated at eight months.
Key Technologies
- Python: Data processing and solution orchestration.
- LLMs — GPT-4 and Claude: Semantic analysis of claims documents.
- AWS SageMaker: Training of custom models.
- FastAPI: Provisioning of service endpoints.
- Power BI: Monitoring dashboards and fraud metrics.
- Databricks: Data pipelines and feature engineering.
Success Factors and Overcome Challenges
- Trust Rebuilt Through Results: Demonstration of measurable impact following six previous pilots that yielded no results.
- Integration with Legacy Systems: Handling multiple data formats present in the claims process.
- Regulatory Compliance: Incorporation of SUSEP requirements into the process automation.
- Demonstrated Economic Viability: Development of a business case approved by senior management within six weeks.
- Validation Before Scaling: Prioritizing a proof of concept (PoC) with real-world impact as the foundation for future investments.
Results
The initiative delivered operational gains and provided a solid foundation for the decision to invest in AI:
- 85% Reduction in Analysis Time: Greater agility in the claims process.
- From 5 Business Days to Less Than 18 Hours: New analysis time per claim.
- Return on Investment Calculated at 8 Months: Financial projection consolidated in the business case.
- Board Approval in 6 Weeks: Feasibility of the initiative recognized by leadership.
- A PoC with Real Impact: Measurable results after six previous unsuccessful pilots.
“We had already tried six AI pilots without success. Foursys delivered real impact in the first PoC—with a business case approved by the board.”
Director of Innovation, Multi-Line Insurance Company