AI in commercial lending: the future of automation and change

AI In commercial lending
AI In commercial lending
AI in commercial lending is here to dominate in a post-pandemic CRE industry

Commercial lending and credit risk assessment are currently undergoing a fundamental transformation with the adoption of Artificial Intelligence (AI) and it is here to stay. With the outbreak of the COVID-19 pandemic small and mid-sized as well as large commercial lenders are beginning to replace manual processes and cross-checks with AI-based quantification and automation. Predictive data analytics generated by AI is not only crucial for better lending decisions, but also for improving efficiency along the supply chain. Hence, AI in commercial lending can be particularly useful for commercial banks, specialty finance companies, and others involved with commercial loans.

What are commercial loans?

Commercial lending is made to organizations or individuals for the sole purpose of business operations, rather than for personal use. It is a type of financing for businesses to increase working capital, improve infrastructure, meet operational costs, etc. Traditionally a manual process, Artificial Intelligence is gradually taking over the commercial lending arena. 

Artificial Intelligence or AI are computer algorithms that imitate human intelligence. Hence, with advanced techniques like machine learning, robotic process automation, Natural language processing, etc, AI holds tremendous potential for automating the overall process of commercial lending. AI in commercial lending simultaneously processes several datasets to automate manual due diligence, credit risk screening, underwriting, fraud detection, etc.

Challenges in traditional Commercial lending

1.     A McKinsey study notes that commercial lending essentially involves a manual commercial underwriting process. 30 to 40 percent of lending resources’ time is spent on manual tasks and scattered systems. The same report observes that the typical “time to decision” for commercial lending is between three and five weeks, while “time to cash,” on average, is three months. As a result, this makes the process inefficient and redundant in terms of credit analysis, loan booking, and portfolio management processes.

2.     Even today, several banks rely on outdated IT systems that prevent the implementation of innovative technology. Moreover, such ancient legacy IT systems increase costs prevent growth and fail to provide a client-friendly experience.

3.     There are also major discrepancies in self-reported data. Banks depend on experienced loan officers to judge the data accuracy and assess the creditworthiness of clients. To identify any doctored data of unrealistic sales and growth projections, these officers seek alternative data sources to clarify the applicant’s ability to repay a loan and scale the size of the loan accordingly. Moreover, the manual involvement elongates the process unnecessarily with chunks of paperwork and a manual underwriting process.

4.  Financial institutions need to comply with the stringent guidelines for commercial lending by regulatory bodies. Hence, it requires them to manage risks from a regulatory, credit, and operational perspective for making better credit decisions. Thus, evaluating and reporting on the risks associated with a loan portfolio incurs huge administrative costs.

5.      Lending institutions often have limited access to data. Hence, it hinders the institutions to gain proper insight into the creditworthiness of the clients. Moreover, such scarcity of data leads to less-than-ideal loan performance management.

6.  The rise of fintech aims to digitize banking operations on a wider scale. Moreover, Fintech seeks to deliver efficiency and a better client experience at lower costs. Further, commercial clients also expect updated services that are powered by technological innovations. With rapid digitization, clients are expecting to close the loans faster with diverse offerings and services, and uninterrupted client experience.

7.  Increased competition from large non-traditional lenders and credit unions necessitates the need for providing a much superior client experience. Hence, the faster and more seamless lending process by non-traditional lenders has driven the need to adopt change.

Use cases of AI in Commercial lending

1. Credit Application Phase

The loan application phase is the origination stage of commercial lending. Financial institutions can use AI to determine credit needs by analyzing credit line usage and evaluating historical data patterns.

With AI, commercial lenders can incorporate several variables to sanction or reject loans to a business.

One crucial factor for businesses is the economic conditions which are usually geography and industry-specific. AI can process large volumes of historical Government data on economic cycles for various geographies and industries. Hence, it creates predictive models that can help to assess the financial projections of the borrower.

By studying how the borrower’s recent financial behavior is different from past behavior, financial lenders can decide whether to move forward with the application and potential causes of concerns. Therefore, early insights enable financial lenders to make faster and effective lending decisions.

2. Underwriting

Financial lending institutions and small and medium-sized banks, credit unions, and online lenders are in need of intuitive, agile banking to provide a better client experience and improve their businesses. Hence, faster processing of applications, fair lending terms, and a transparent underwriting process.

The primary goal of an AI-based underwriting process is to analyze and assess the borrowers who qualify for approval and loan terms by default. Commercial loan underwriting has always inherently been a manual process with humans analyzing financial aspects like cash flow forecasts and debt coverage ratios.

However, such manual underwriting is giving way to automated credit assessment. Predictive models powered by AI can create models based on the borrower’s historical loan performance. Such predictive analysis can automatically assess credit risk and infer similar results as the standard analysis by loan officers.

Moreover, AI excels at identifying low-risk borrowers much easier than traditional underwriting techniques by a wider variety of credit variables in the predictive model.  As a result, the volume of approval rates increases by almost 10%. Further, AI helps in achieving consistent risk assessment and accurate decision-making regarding consumer lending.

3. Data Gathering

Traditional modeling techniques rely on financial statements and qualitative information of a consumer.

While financial statements comprise ratios for liquidity and coverage, qualitative metrics include the assessment of the management team’s efficiency, the firm’s competitive position or appreciation of the firm’s physical location, etc.

However, the collection of such data is both time-consuming and costly. However, AI algorithms assess such non-numerical factors to evaluate the creditworthiness of an applicant. Natural Language Processing (NLP) can assess social media feeds to gain insight into their social media activity.

For instance, negative reviews for a hotel or restaurant can affect credit risk. Besides assessing financial statements, and monitoring consumer behavior on social media, AI algorithms can extract transactional data as well. Examples include a history of late payments, cash flow movements, line of credit usage, patterns of deposits, etc. Hence, incorporating diversified data sources can optimize the commercial lending process.

4. Credit Risk Quantification

Credit risk management continues to be a major problem for financial institutions due to inefficient data management. Unintuitive visualization into the borrower/s ability to pay back and lack of risk assessment tools.

AI algorithms have the potential to process the exponential volume of available data to impact credit risk management. AI can extract insights from unstructured alternate data sources to verify the credibility of the information provided by the applicants.

Moreover, traditional probability of default (PD) models depend heavily on logistic regression and do not leverage the predictive power in the data. Organizations can use this data to build better loss-given default (LGD) models. Further, AI can be used to discover early patterns of warning signals from various sources with advanced computation.

However, regulatory compliance, data quality, and model governance still challenge the implementation of AI for credit risk quantification. Nevertheless, the advanced algorithms of AI in commercial lending can optimize the management of credit risk.

5. Pricing ad RAROC

AI and Machine learning algorithms can be used in pricing and risk-adjusted returns on capital (RAROC). Deep learning methods such as artificial neural networks to estimate the economic capital consumption of the loans. (Artificial neuron networks implement the mathematical representation of a biological neuron network that can manage non-linear and interactive patterns between variables. Hence, this can be particularly useful for company credit scoring.)

Hence, the need to fully re-run the lengthy economic modeling process is eliminated. This, in turn, leads to better pricing decisions. In addition to this, pricing models can be more precise if they include more dynamics other than risk-related ones. Qualitative dynamics such as the extent of the relationship, long-term prospect of industry, growth potential, etc help in the efficient structuring of pricing models.

6. Legal compliance

Since commercial lending is a complex legal aspect, it generates a huge chunk of paperwork to execute the transaction. It also secures the lender’s interest in the collateral.

With advanced algorithms, AI-based Machine learning can examine data to identify discrepancies in the data which might go undetected in manual 5. processing. These algorithms surpass simple rule-based checks and can identify exceptions that are out of line with comparable deals.

Hence, AI can handle the redundant tasks of validating documents which allows lenders to save money and ensure that documents are legally compliant.

7. Back-end management

AI can re-engineer back-end processes for banks. For instance, business institutions need to disclose and submit their current state of accounts and other assets to the banks at certain intervals. This is done to assess the financial health of the company.

Moreover, this is also done to ensure that the business cannot borrow any further if the previous thresholds aren’t met.  However, these evaluations are labor-intensive and require manual processing. Hence, AI tools combined with robotic process automation (RPA) make the process more efficient and prompt. 

8. Fraud detection

Loan stacking means customers taking several loans from multiple lenders. This is considered a serious cybercrime and it is a common phenomenon in the lending business. However, AI in commercial lending helps in monitoring the suspicious and unusual behavioral patterns of the loan applicant.

Hence, AI can identify fraudulent practices in lending. For instance, in case a person already has stacked various loans from multiple lenders, AI can detect this and thereby flag him/her from further loan stacking.

How does AI help commercial lenders and borrower?

The Lilypads Bottomline

Commercial lending in traditional circumstances relied heavily on judgemental analysis by loan officers and less on predictive analytics. However, the implementation of AI in the commercial lending industry can help attract massive traction and benefits. Such programs can help improve underwriting, accuracy, and efficiency. Moreover, AI mitigates risks and errors from the lender’s perspective. In addition to this, borrowers can improve from faster loan approvals and enhanced customer satisfaction. Hence, AI in commercial lending automates and foolproofs the entire lending process from loan origination to management procedure.