Objective

Each of the issue description in the risk management network was compared against each others to

Compute the similarity of the descriptions
Flag the descriptions that has a similarity score above a specific threshold for further review and analysis
Ability to compare the issue similarity within / between process domains
Ability to identify domains that has higher risk

Industry: Financial Services – Banking

Services: Technology Innovation

Technology: ML/NLP Analytics

Enabled to identify the problem areas under the process domains that contributed to risks.

Enabled the solution for a problem that was nearly impossible to solve based on human talent alone due to volumetrics.

Accelerated QA efforts as the real-time execution and removed the human dependency.

Our Client

Our client, a Fortune 100 company, is a global leader in the financial services sector, renowned for its diverse range of banking and investment products. With a strong presence in numerous countries, it offers innovative solutions for consumers, corporations, governments, and institutions worldwide.

Challenge

As part of the process efficiency improvement efforts to address Risk, the need was to identify and correlate various issues/complaints under the process domains and the impact on the Risk mitigation process.

Approach

Employ a Natural Language Based (NLP) system to map the issues to pre-defined classes

Designed an NLP-based multi-class model for mapping the issues.
Develop a system with a GUI-based front end for batch and on-demand execution of the model.

Outcomes

The software solution enabled the comparison of issues
Ability to identify the problem domains and flag for further scrutiny
Ability to identify similar issues across the process domains
Enabled self-service dashboards for ad-hoc analysis for batch processing and historical reporting

Impact

Reduction in human capacity requirements from thousands of hours to a day to run the model and validate
Reduced the time to market from Weeks to hours
Increased the efficiency of the solution from being able to manually deliver a specific number of batches per month to real-time mapping on streaming data