Automating Health Insurance RFP Processing with Deep Learning

The Challenge

 Acquiring valued and profitable business is the goal for insurance carriers, but this involves participating in various Request for Proposals (RFPs). Insurance firms are required to evaluate RFPs and respond with quotes. The RFPs involve a voluminous amount of documents and information, including requirements, questionnaires, coverage, and benefits expected by the requestor. There are three main problem areas:

• Manual processing of documents sent by the requestor to understand their expectations.
• Providers need to base quotes on census information about their employees and their background. Different brokers have their own ways of representing this information, such as employee dob, birth date, emp dob for date of birth. It is a challenge to identify these variations and process them.
• Normalizing the data to standard formats.

The Solution

We used deep learning techniques to automate the process of going through documents and extracting vital information required to provide quotes. To address the variations and different formats in the census, natural language inference techniques were used to convert various columns representing the same information to a standard format. This allowed us to clean and process the data efficiently.

The Outcome

Clients relied heavily on manual processes to decipher and process these documents, which typically took several weeks to complete, especially when it came to census files. However, with the introduction of automation, we achieved a remarkable transformation, enabling us to process each census file in just 3 minutes.

Tech Stack:

• Python
• PyTorch, TensorFlow
• Pandas and NumPy