Case Studies Top 3 US bank Leverages AI and NLP to streamline financial document processing

Edit This Case Study Record

	Top 3 US bank Leverages AI and NLP to streamline financial document processing

Top 3 US bank Leverages AI and NLP to streamline financial document processing

Technology Category	Analytics & Modeling - Machine Learning Analytics & Modeling - Natural Language Processing (NLP) Application Infrastructure & Middleware - Data Exchange & Integration
Applicable Industries	Finance & Insurance
Applicable Functions	Business Operation Quality Assurance
Services	Software Design & Engineering Services System Integration Training
Challenge	Analysts at this top 3 US bank spend hundreds of hours a year manually reviewing financial documents to find information on interest rate swaps. This manual process is time-consuming and takes away from their ability to assist customers proactively. The team recognized the potential of using AI and NLP to streamline 10-K processing but lacked the training data required to train a model that could automatically identify and extract interest rate swaps from 10-Ks accurately across multiple formats. Read More
About Customer	The customer is a top 3 US bank, which is a major financial institution with a large team of financial analysts. These analysts are responsible for reviewing financial documents, such as 10-Ks, to find specific information like interest rate swaps. The bank is focused on improving operational efficiency and leveraging advanced technologies to enhance their services. With a significant number of employees and a vast amount of financial data to process, the bank is looking for innovative solutions to reduce manual workload and improve accuracy in data extraction. Read More
Solution	The bank leveraged programmatic labeling and weak supervision to encode analyst expertise as labeling functions (LFs). This approach allowed them to train a custom NLP model that could automatically identify and extract interest rate swaps from 10-Ks. The model achieved an F1 score of 83 in just a few weeks. By using Snorkel Flow, the team was able to generate 70,000 labels per minute programmatically, significantly speeding up the training process. This solution not only reduced the time spent on manual document review but also improved the accuracy and consistency of the extracted data. Read More Log in to view content
Contents

Technology Category

Analytics & Modeling - Machine Learning

Analytics & Modeling - Natural Language Processing (NLP)

Application Infrastructure & Middleware - Data Exchange & Integration

Applicable Industries

Finance & Insurance

Applicable Functions

Business Operation

Quality Assurance

Services

Software Design & Engineering Services

System Integration

Training

Challenge

Analysts at this top 3 US bank spend hundreds of hours a year manually reviewing financial documents to find information on interest rate swaps. This manual process is time-consuming and takes away from their ability to assist customers proactively. The team recognized the potential of using AI and NLP to streamline 10-K processing but lacked the training data required to train a model that could automatically identify and extract interest rate swaps from 10-Ks accurately across multiple formats.

About Customer

The customer is a top 3 US bank, which is a major financial institution with a large team of financial analysts. These analysts are responsible for reviewing financial documents, such as 10-Ks, to find specific information like interest rate swaps. The bank is focused on improving operational efficiency and leveraging advanced technologies to enhance their services. With a significant number of employees and a vast amount of financial data to process, the bank is looking for innovative solutions to reduce manual workload and improve accuracy in data extraction.

Solution

The bank leveraged programmatic labeling and weak supervision to encode analyst expertise as labeling functions (LFs). This approach allowed them to train a custom NLP model that could automatically identify and extract interest rate swaps from 10-Ks. The model achieved an F1 score of 83 in just a few weeks. By using Snorkel Flow, the team was able to generate 70,000 labels per minute programmatically, significantly speeding up the training process. This solution not only reduced the time spent on manual document review but also improved the accuracy and consistency of the extracted data.

Impact #1	The implementation of the custom NLP model saved over 2000 hours per year for financial analysts, allowing them to focus on more value-added tasks.
Impact #2	The use of Snorkel Flow enabled the team to generate labels programmatically at a rate of 70,000 labels per minute, drastically reducing the time required for model training.
Impact #3	The solution was developed and deployed in just six weeks, demonstrating the efficiency and effectiveness of the approach.

Impact #1

The implementation of the custom NLP model saved over 2000 hours per year for financial analysts, allowing them to focus on more value-added tasks.

Impact #2

The use of Snorkel Flow enabled the team to generate labels programmatically at a rate of 70,000 labels per minute, drastically reducing the time required for model training.

Impact #3

The solution was developed and deployed in just six weeks, demonstrating the efficiency and effectiveness of the approach.

Benefit #1	2000+ hours per year saved for financial analysts.
Benefit #2	6 weeks to build a production-quality AI application.
Benefit #3	70,000 labels per minute programmatically generated via Snorkel Flow.

Benefit #1

2000+ hours per year saved for financial analysts.

Benefit #2

6 weeks to build a production-quality AI application.

Benefit #3

70,000 labels per minute programmatically generated via Snorkel Flow.

Download PDF Version

Overview

Top 3 US bank Leverages AI and NLP to streamline financial document processing

Operational Impact

Quantitative Benefit