In an effort to keep up with the changing market dynamics, banking and financial institutions are accelerating their investment in digital transformation initiatives. Their efforts are aimed at delivering a greater customer experience at a lower cost while fending off competition from the extremely responsive, digitally native fin-tech start-ups.

Most digital transformation initiatives are centred on the automation of end-to-end credit processes. The focus is on credit, as it forms the foundation of most customer relationships. Getting the credit transformation correct is expected to deliver a seamless experience, enable a faster lead time and lower customer processing costs — all of which will yield a variety of benefits.

The ongoing digitisation efforts of credit workflows have achieved varied levels of transformation and automation across diverse segments, including personal, retail, mortgage and corporate. A few personal loan product disbursements are clocking a cycle time of minutes and are delivering the coveted seamless experience, while other unstructured content-based processes, such as mortgage loan processing, are left far behind. These processes are riddled with multiple challenges to deliver transformative experiences,  such as disbursement, in less than eight hours, i.e., one  business day.

Despite the digital transformation drive, many loan applications and supporting documents, such as proof of income, pay stubs, W-2 forms, legal contracts and mortgage contracts, are paper based and unstructured in nature. A lot of effort is invested in processing this content. Lenders must manually validate and classify documents, extract relevant information, reference supporting documents, request missing content, and ultimately enter relevant information into loan management systems for downstream credit scoring and analysis. 

More on what Intelligent Automation is and what it can do for your business to come in future blog posts! Check out past blogs here.

This continued reliance on manual processing results in a slower loan origination cycle time, errors leading to likely faulty underwriting and challenges in compliance to regulatory norms.  To overcome these challenges and competitively drive transformational results, financial institutions must leverage artificial intelligence (AI) to intelligently automate the end-to-end credit process.

To achieve true end to end automation, AI can be leveraged for:

  • Automation of traditionally slow and often paper-based/unstructured data processing with minimal human intervention.
  • Utilisation of enterprise data to automate decision- making to deliver decision intelligence.

For automating unstructured data processing, computer vision can be used in coordination with natural language processing (NLP) to convert the digitally scanned document images/PDFs into structured, machine-readable data. This transformation enables running various AI and ML models to promote automated decision-making. This is rendered possible by the application of various NLP models, text analytics models, domain vocabularies, ontologies, and proprietary knowledge bases, trained via unsupervised ML models etc., to reach a point of structure.

This use of AI completely changes the nature of document processing. The lender that earlier was responsible to manually capture data now is guided through the review and handling of exceptions. In the case of any adjustments or corrections being required, as identified by the low confidence score related to the extraction models, the lender reviewing the output can make the required changes. These changes captured by the system can be systematically fed into ML feedback loops in order to continuously train the system  to achieve greater accuracy.

However, applying AI to extract and structure information is only the first step. Predictive models, text analysis and comprehension (NLP, etc.) AI agents can be used to analyse the customer information and credit history and provide suggestions to credit analysts on loan approvals and associated risks. ML algorithms can also be used to further curate and grow the lender’s proprietary knowledge base on customer information and credit history, such that variations and exceptions handling can be taught to the system.

Furthermore, after the loan is approved, the communication of the commitment can be automated using natural language- generation algorithms that are based on deterministic or algorithmic rules, and disbursement can be automatically initiated in order to close the end-to-end automation loop.  

The proper execution and implementation of AI can deliver a superior customer experience and help drive toward disbursements of paper-based processes within one business day, i.e., eight hours.