Case Study - NLP BFSI

NLP Based Insurance Claim Communication Automation and Analytics

MQI team has designed solution for one of the NLP (Natural Language Processing) project for one of the clients in BFSI Sector. The solution designed to reduce manual intervention, quick turnaround and improve customer satisfaction in claim processing module.

Problem statement at a glance

There are several emails for insurance claim settlements coming to any travel company and/or Insurance Company. First issue is, these mails are processed manually as they are written by variety of customers in free flow format. It takes very long to identify relevant information, find the missing information needed to process the claim and revert to customer asking additional details. Second part is to submit the completed forms to processing team in timely manner and third is to keep customer updated on progress.

Prime Objective of the assignment was to reduces the claim settlement period and automate the manual effort that is involved in settling the claims.

The policyholder sends an email reporting a complained, this email needs to be classified into a category and relevant information to settle the claim needs to be aggregated from the email, insurance provider etc. This whole process needs to be streamlined also in the process enhancing the quality of customer engagement.

Proposed Flow and Approach

slider image

Proposed Approach for End to End Automations

Phase I

Client receives (complaint) email from customers

  • The NLP engine will receive the details for classification on the type of insurance
  • NLP engine will work on the entities to be extracted from the text & attachments
  • Using the extracted fields populate the form
  • Identify the missing entries and enlist those in the form (this form will be hosted on the server)

Process Steps

  • Client receives (complaint) email from customers
  • Classification on the type of insurance is done -AI
  • Entities need to be extracted from the text-AI
  • Entities need to be extracted from the attachments-AI
  • Fill up the form fields or match them-AI
  • Identify the missing entries and enlist-AI

Phase II

  • Send the e-mail to the customer along with the link of this form (send reminders)
  • Allow the user to complete filling the form & submit the same
  • Validate the form using the NLP engineer
  • Verify the form manual and forward it to the SAX database.

Phase III

  • Calling and dial in feature using NLP or a Chatbot enabled feature for extracting accurate information.
  • Dial the customers and engage with them on a dialogue to extract the missing information
  • This would eliminate the need of e-mail reminders etc. and would be an immense value add to AGC.

Note : Phases proposed as per client expectation, however, to reduce timelines , entire assignment can be taken up together and stagewise release can be planned.


1. Screen details

  • Web interface: For testing first cut, an i/f provision ded for the user to paste the contents of the email.
  • We can automate it as a batch in later stages

2. Mail contents populated

Mail Contents Populated



Component Details

  • NLP Engine: The Natural language processing engine understands the email text content needed for the specific type of claim and classify the category or the type of service the customer is asking for and to analyses the corresponding text. The NLP engine will perform a NER (Named Entity Recognition), this component will be trained and learn from existing data. The NLP engine will also include Classification, and the parsing of the extracted entity.
  • User interface: A simple web based i/f (or as per client requirement)
  • Access to API’s which have information related to the type of Insurance, each insurance details etc.
  • Email facility to send auto-emails etc. to the customer.

Thank You