Claims IDP
Classification and summarization of multi-page documents using OCR and AI techniques to improve the efficiency and accuracy of claims document processing.
Key Metrics
Target Users
- Claims Adjusters
- Claims Managers
- Underwriters
- Operations Leaders
Business Benefits
- Enhanced operational efficiency
- Reduced processing time
- Improved accuracy and consistency
- Better resource allocation
How It Works
How it works:
1. When a new claim document is uploaded, OCR extracts the text while the system analyzes both text and visual elements to determine document type.
2. Three different approaches were validated:
- BERT+ResNet: Combined NLP and image analysis (70% accuracy)
- LayoutLMv2: Transformer model incorporating document structure (100% accuracy, given the right data)
- Agentic LLM workflow: Multi-agent setup with task breakdown (90% accuracy, highly scalable)
3. The system extracts relevant information including policyholder details, claim numbers, and document content using either machine learning or LLM approaches, depending on the document type.
4. For multi-page documents, similar pages are clustered to optimize processing time and improve results.
Business Benefits:
- 50%+ reduction in document processing time
- 90%+ classification accuracy
- Consistent handling of claims across the organization
- Better resource allocation based on claim complexity and priority
- Improved customer experience through faster claims resolution
Operational Impact
- Streamlined workflows
- Accelerated claims processing
- Reduced manual data entry
Financial Impact
- Lower operational costs
- Decreased error-related expenses
- Improved resource allocation
Customer Impact
- Faster claims resolution
- Enhanced customer satisfaction
- Reduced follow-up inquiries
Technical Highlights
Implementation Timeline
No documents uploaded yet
Business Context
In the insurance industry, the claims process involves handling a vast number of multi-page documents, which are often scanned images. These documents need to be classified correctly to ensure they are processed efficiently. Manual classification is time-consuming, error-prone, and can lead to delays in claims processing. Automating this process with AI could significantly enhance operational efficiency and reduce errors.
Problem Statement
The current manual classification of multi-page insurance claims documents is inefficient and prone to errors. This PoC aims to solve this problem by developing an automated classification system using a combination of OCR and ML models/GenAI.
Impact and Importance
Solving this problem will result in reduced processing times, lower operational costs, and improved customer satisfaction due to faster claims handling. Additionally, it will allow human resources to focus on more complex tasks that require critical thinking.
Developer Setup
To set up and run this PoC locally, follow these steps:
- Ensure you have Python 3.9+ and Quart installed on your system.
- Clone the repository containing the PoC code. Navigate to the 'app' folder and install the dependencies:
pip install -r requirements.txt - Ensure the Quart service is running on
localhost:11001by executing:python main.py - Upload documents via the PoC front end for classification.
How to Use This PoC
Follow these steps to use the PoC:
- Click the "Upload" button and select a document (.tif, .pdf).
- Disable summarizer if not needed.
- Wait for the Quart service to process upto 4 files in parallel.
- The table will be populated with a view action as the response is received from Quart event stream.
- Click view button for more details extracted from the document.