• Customer is a Global Commercial Bank, with operations ranging from retail, commercial, to mortgaging
• In the highly competitive domain of “Bill Discounting” or “Factoring”, customer wanted a system that could automate and provide data points to arrive at a decision on offering credit discounts to their
• SME (Small, Medium Enterprise) clients SME clients will share their receivable statements, invoice summaries and stock statements in many formats

Introduction

Customer is a Global Commercial Bank, with operations ranging from retail, commercial, to mortgaging

Challenges

• An in depth study of input documents was done to determine a methodology of forming patterns
• Training data set, test data set and validation data set were created
• Overall process was built in coordination with the customer to make sure that human intervention is sought at appropriate step to augment AI
• A hybrid combination of human intervention and AI/ML ensured that the outcome is accurate and complete

Solution

• Input documents were divided into scanned and regular PDFs and accordingly data extraction mechanisms were deployed
• Based on the nature and complexity of data, appropriate business rules and ML algorithms were identified to extract right data from the stream of input documents
• Output format was in JSON and it was tailored to contain only specific data points
• Horizontally scalable architecture was adopted

Key Benefits

• In a competitive market of credit processing, the bank acquired an edge to process SME requests rapidly without compromising on quality. A competitive edge
• Combination of human intervention and AI made sure that the risk is minimized when it came to critical decisions on credit line processing. Minimized risk
• AI models were further applied on the data gathered over a period to identify trends and predict decisions on credit processing