Industry
Banking
Banking Services Transformation: Unlocking Exponential Value
Highlights
- Timely access to funds is critical for business organizations to consistently sustain operations and grab growth opportunities as they arise.
- However, the process for availing credit facilities from financial institutions is cumbersome and time-consuming, frequently resulting in missed business opportunities and loss of market primacy.
- To retain customers and grow business, financial institutions must incorporate GenAI into the various stages in the loan journey, especially the credit memo preparation step, which is critical to credit decisions.
Overview
For businesses, access to capital is crucial for pursuing growth opportunities and managing operational needs.
While customers’ need for a business loan is often urgent, the pace of approvals is slow due to extensive paperwork, lengthy procedures, and slow approval cycles. For customers, this can translate into missing out on time-sensitive business opportunities, adverse impact on cash flow, and loss of market share. For banks, this can result in customer churn and a smaller loan book, adversely affecting profitability and competitiveness. It can also pave the way for agile, digital fintech startups as well as big tech to enter this space, exposing traditional banks to disruption.
Let us examine the existing loan process followed by incumbents—traditional banks traverse multiple stages (see Figure 1) while sanctioning a business loan, ranging from gathering client documents, to underwriting and loan structuring, right until disbursement and monitoring. Each stage plays a critical role in evaluating borrowers’ eligibility and risk.
Figure 1: The commercial lending journey highlighting where the credit memo fits
Clearly, overcoming the challenges in the loan approval process is a critical imperative for commercial banks to retain an edge. The solution lies in adopting generative AI (GenAI) technologies across the business loan approval process, enhancing efficiency, reducing costs, and improving borrower experience as well as profitability. While GenAI has the potential to reimagine every stage of the loan approval process, banks must prioritize GenAI adoption in credit memo preparation, given that it forms the basis of loan approval or rejection decisions.
Challenges in credit memo preparation
In the first stage of a new loan request, banks prepare a credit note or a credit memo.
This is a key pre-underwriting primary evaluation of borrowers, comprising both qualitative and quantitative analyses. Credit memo preparation is a manual, time-consuming, and error-prone process, requiring credit officers to refer to multiple data sources such as customer relationship management (CRM) systems, borrowers’ financial statements, tax returns, forecasting models, registration and incorporation certificates, ownership structure, asset valuation report, and credit reports, among others. It also requires them to use financial spreading tools and adhere to regulatory requirements.
Credit officers typically spend several hours collating and analyzing data for each individual borrower. Each credit application spans 15 to 20 pages and consists of different sections on borrower profile and ownership, business model, financial analysis, risk, policy exceptions, pricing and fees, group relationships, and so on. Consequently, credit memo preparation significantly delays the business loan approval process, resulting in adverse business impact and poor borrower experience. The way forward lies in using GenAI to automate the credit memo preparation process, right from summarizing borrowers’ information to analyzing potential risks.
Enter GenAI
GenAI tools can be deployed across multiple steps in the credit memo generation process, expediting loan approvals (see Figure 2).
Figure 2: GenAI for automating credit memo generation
Borrower information has to be gathered from a variety of documents including bank statements, trade certificates and annual reports, among others. This is currently time-consuming and error prone as it is a manual process. A combination of GenAI and optical character recognition (OCR) tools can be used to reduce cycle time and errors.
In addition, bolstering borrower data with background information available from trusted public sources can help create a detailed picture of borrowers. One way is to check for any adverse coverage of prospective borrowers or their management teams in media sites. Similarly, banks must scan the website and annual reports filed with regulators to understand more about the borrowers’ business, the industry they operate in, and the products they offer. Banks must verify borrowers’ ownership profile by scrutinizing information available in trade registries and credit bureaus, evaluate analyst reports and information available on platforms like IBISWorld and Bloomberg, and analyze their business models to flag potential risks. While GenAI technologies can help streamline and speed up all these tasks, they may not always be 100 percent accurate, in case the large language models (LLMs) are trained with open or public data. Banks must take care to ensure that the LLM model delivers accurate output.
A credit memo must offer a detailed analysis and summary of the financial health and performance of borrowers for current as well as previous years based on various documents. GenAI tools can assist in analyzing the documents, extracting key numbers, calculating ratios, and generating a brief commentary on the financial position of the borrower by connecting the dots between the numbers in financial statements and the notes to annual reports. For example, GenAI can help in determining if the average collection period has increased and the causes such as cash flow problems or piling up of inventories. GenAI flags items for which the credit officer must seek detailed clarifications.
Financial analysis is predominantly a subjective exercise and not just number crunching. Drawing the right inferences from numbers requires the specialized skills of an experienced credit officer. While GenAI can automate the extraction of key numbers from financial statements and calculate ratios, it may not have the required ability to draw nuanced inferences. This is because several market- and borrower-specific factors that influence business performance are neither mentioned in annual reports nor can they be accurately interpreted. Moreover, financial statements are often subject to window-dressing. Thus, evaluating all these elements to arrive at the right credit decision demands human judgement and reasoning.
To mitigate credit risk, financial institutions need to assess the quality and value of the collateral pledged or the asset for which the loan is requested. The objective is to determine if the asset can be easily liquidated to recover the loan in case of default, thereby protecting the bank’s investment. This involves wading through voluminous and complex data. For example, in commercial real estate, risk assessment is based on data in the property appraisal reports which can run up to ~100 pages with detailed information on asset quality, neighborhood, prices at which comparable properties have been sold, as well as inspection reports and so on. Deploying GenAI tools can simplify and speed up the extraction of collateral related information and summarize it, to reduce the time taken to generate credit memos.
Banks typically create a risk profile of the borrower as part of risk assessment. This exercise involves analyzing the financial institution’s credit policy to determine eligibility and flag risks. Credit policies are usually comprehensive and comprise several documents on policy exception, fee structure, and loan terms and conditions, among others. Banks can deploy LLMs to intelligently recommend the relevant terms and conditions that will apply to a particular borrower. While this may seem like a fairly simple rule-based exercise, it actually requires a thorough analysis of multiple documents. Credit officers are usually well versed with the bank’s policies and can recall the type of covenants recommended for similar loans sanctioned in the past. But an LLM model will need to be trained comprehensively to draw the right inferences.
Once the credit memo is ready, GenAI can be leveraged to perform a SWOT analysis, flag potential risks, and create a holistic view of the borrower. This will aid the credit committee in arriving at an informed loan decision.
Clearly, GenAI offers the potential to boost the operational efficiency of the credit approval and decision-making process across the pre-underwriting and underwriting stages. However, developing an end-to-end solution for automating the business loan approval process is a complex, long-term exercise. Banks must select the right use cases, keeping in mind regulatory requirements as well as risks around privacy, security, explainability, and algorithmic bias, while prioritizing human-in-the-loop use cases to enhance accuracy.
Furthermore, banks will have to be mindful that certain GenAI use cases may demand modifications to existing internal systems and processes, especially around adaptability and readiness in the data management space. The way forward is to embrace a ‘fail fast’ approach, running pilot programs to develop a minimum viable product (MVP) for a specific loan segment, and then scale across other segments.
Looking ahead
The ability to adapt will become a prerequisite to thriving in a dynamic commercial lending environment.
Leveraging GenAI to automate lending processes is more than a tactical efficiency play, it is a strategic investment aimed at future-proofing for the challenges and opportunities of tomorrow. To realize the long-term benefits of GenAI in lending, banks must move beyond pilot projects and commit to a culture of perpetual adaptation—where technology, people, and processes evolve in tandem.
Banks must act quickly—those that do will gallop ahead of their peers, setting new benchmarks in customer value and operational excellence. However, the journey to unleashing the transformative potential of AI in the lending function will come with its own pitfalls. Given the complexities involved, banks must consider partnering with a service provider with the requisite domain knowledge and technology expertise for hassle-free implementation.