Payslip, bank statement analysis

AI-powered data extraction from payslips and bank statements

The business problem

Loan application processes entail the evaluation of customers' financial stature. To arrive at judicious lending decisions, the inputs used must be credible and accurate/definitive/categorical.

The most common resources are the applicant's payslips and bank statements. Approval workflows start with the identification and extraction of relevant details from such documents. This is a pivotal step--successive processes rely on this data to determine the customer's creditworthiness, indicated by parameters such as the income-to-expense ratio.

Traditional loan processing tasks depend heavily on manual efforts. The analysis of the documents, compilation and correlation of data, and computation of parameters for the final decision are time-consuming, laborious, and error-prone, besides being vulnerable to subjectivity/bias.

Customers' salary slips and bank account statements are in diverse file formats such as PDF, images, or HTML.

The terminology used for the descriptions in the documents varies among organizations and banks.

There is a wide variation in the document layout, with the information presented in different sequences, pages, or locations within a page.

Quantrium's solution

A leading Asian bank with a substantial market share in the personal loan segment was keen to speed up approval workflows. Manual processes were found to be inadequate as the volumes and complexity of the documents grew. Scaling up of the resources warranted additional OPEX with disproportionately poor returns. Quantrium studied the requirement and developed a platform for automated scrutiny of payslips and bank statements. The AI-powered solution accurately recognizes over 1,000 document formats and the data contained in them across diverse file types and layout styles. The essential details are categorized and the bank's decision-support systems/database updated over an API instantaneously.

Extensively trained with thousands of real-world documents, the solution facilitates a robust, risk-free loan application process. Quantrium continues to enhance its capabilities to identify logos, correlate data across documents, check data integrity, apply personalized rule engines, and support multilingual documents.

Customers' salary slips and bank account statements are in diverse file formats such as PDF, images, or HTML.

The terminology used for the descriptions in the documents varies among organizations and banks.

There is a wide variation in the document layout, with the information presented in different sequences, pages, or locations within a page.

Text and Image Extraction

Advanced text recognition and visual analytic techniques accurately detect the content in documents, regardless of their format, structure, and size. Advanced image pre-processing to reduce noise, correct skewness etc., in images. The solution effortlessly handles multi-currency and multi-page documents. Custom-built mapper algorithms rationalize the labels and associate the data with appropriate fields.

Intelligent Grouping

Leveraging NLP/ML, the solution classifies the collected data into logical groups for further analysis. Incomes and expenses are dissected into subcategories, facilitating an accurate interpretation of the customer's earning potential and spending patterns.

Verification

The solution checks the data integrity within and across documents. Integration with government and public data repositories allows fail-proof validation of the information. The logo in salary slips or bank statements is also verified for authenticity.

Intelligent Personalized Rule Engine

A personalized rule engine augments the solution's capability to meet varied needs. Assessment methodologies can be fine-tuned to incorporate customer- and product-specific parameters and metrics.

Implementation

Quantrium's specialists developed the solution with custom-built AI/ML algorithms and models. Thousands of documents were used for training and perfecting the design. Quantrium's cloud engineering team optimized the architecture to offer the scalability and performance required while ensuring compliance with regulatory and infosec mandates. A contemporary user interface ensured quick and trouble-free adoption.

Customer Experience

Quantrium's solution powered a remarkable transformation of the client's loan origination processes. Today, the evaluation of customers' documents gets completed in seconds, enabling the bank to claim the top spot for the quickest approval times in the industry. Human efforts are minimized, credit risks lowered, and processing efficiency enhanced, leading to substantial cost savings and business growth.

Capability to fulfill business goals:
= 100%

Data Accuracy
> 99%

Infographic document processing time
< 60 seconds

Retail loans are important products in our portfolio. Our business objectives required a faster, reliable technique to process loan applications. Quantrium's AI solution fulfilled our needs perfectly. We clear more applications in a shorter time with near-100% precision, and our customers and internal stakeholders are delighted with the results.

- CIO of the bank

Technologies Used

PDF to Text extraction

Object Recognition

Identification

Object Detection

AI tools

Deep Learning

Decision Tree