A Fortune 500 Insurance company needed a solution to streamline downstream processes on claim platforms to improve employee productivity and increase margins. The solution needed to easily ingest, store, and analyze data from diverse sets of structured and unstructured data.
Client Challenges and Requirements
- Errors and inconsistencies in the claims platforms.
- Inefficiencies in claims processing.
- Data quality issues.
- Need for a scalable, accurate, and efficient solution to extract key data points from diverse data sources.
Bitwise Solution
- Bitwise developed an artificial intelligence (AI) solution that utilizes Large Language Model (LLM) to process and understand the context of unstructured documents. The solution analyzes claims data by automatically reading insurance documents and extracting the important information.
The system uses Databricks to build a 3-step data analysis process:
- Bronze Layer – securely stores raw data of all information from policy documents (such as emails with attachments).
- Silver Layer – smart AI-powered reading assistant analyzes the raw data then extracts and holds key details like coverage amounts, deductibles, and expiry dates.
- Gold Layer – once a human analyst double-checks the extracted information for accuracy, the validated, clean data goes into the Gold Layer, which is the trusted source for everything insurance related.
Tools & Technologies We Used
Azure Databricks
DBRX Instruct LLM
LLAMA3 8B Model
PySpark
HuggingFace
Key Results

90% faster Quote turnaround with reduced TAT from 2 days to 2 hours

Increased productivity resulting in accelerated time-to-insight

85% accuracy in automated Data Extraction with LLAMA3 8B

Enhanced scalability to easily handle large volumes of unstructured data