Credit provision is a key driver of economic growth. But despite stringent regulations and strong fundamentals, the Indian economy suffers from a severe credit gap. A good indicator of this gap is the credit-to-gross domestic product (GDP) ratio, which is 50% in India compared to 177% in China. The impact of this gap is severe for micro, small and medium enterprises (MSMEs) and nano-SME borrowers, as the current banking infrastructure does not adequately reach them due to high operational costs and underwriting challenges. Herein lie the most impactful opportunities for artificial intelligence (AI) and machine learning (ML) in credit provision and decision-making.
ICRA estimates that credit is expected to grow by 16% in FY24, driven by demand for small unsecured loans. While this growth is healthy, growing concerns over poor lending practices, including over-indebtedness and substandard underwriting, have led the regulator (Reserve Bank of India) to tighten lending standards. This tightening is likely to slow credit growth to 11-12% in FY25, highlighting the importance of risk management in small loans, i.e. very low-cost lending.
To understand and measure risk, or a borrower's creditworthiness, two things need to be assessed: ability and willingness to repay.
AI models provide a versatile toolkit for different stages of the customer lifecycle within financial institutions. These applications broadly fall into a few categories:
Credit decisioning: The adoption of AI/ML techniques in credit decisioning requires the use of supervised or unsupervised learning algorithms. For example, ML can be used to analyze credit bureau reports to uncover insights into misreported loans, specific repayment structures like lump sums, default trends across different regions and professions, and income distribution within districts and states. Such analysis can help gauge a user's repayment capability.
· Detecting fraud and bad actors: Potential red flags can be identified by scrutinizing user behavior during loan applications, interactions with the application, copy-paste tendencies, frequency of data modifications, connection changes, etc. On the KYC front, assessing the integrity of user data across various sources can help spot fraudulent borrowers and assess their willingness to repay.
Early warning signs: After a loan is disbursed, financial institutions need to closely monitor repayment patterns. By scrutinizing credit bureau data and employing ML techniques, they can identify risks and take proactive measures for successful collection.
Operational Efficiency: Intelligent systems can streamline operational workflows by learning and automating actions typically performed by operations teams. Implementation of ML techniques significantly reduces turnaround time (TAT) and minimizes error rates due to manual intervention.
Improve collection efficiency: For lending institutions, effective collections are paramount. AI models can identify repayment patterns, preferred repayment methods, and user communication interactions, allowing them to proactively resolve collection issues.
Choosing the right AI/ML algorithm depends on the nature of the business and the quality of the collected data. For institutions working with unstructured data, unsupervised learning provides valuable insights. Clustering or association algorithms are viable choices for generating models in this context. Conversely, supervised learning is more suitable for established financial institutions that leverage collective intelligence from user data. Regression and classification are the main algorithm types used in such models.
Two credit subsectors are expected to see significant AI-related adoption in the coming years. The first is female borrowers, who are outpacing men in demand for credit, especially for SME loans. While female borrowers typically have less traditional underwriting data available at the time of application, they have ample alternative data in the form of savings + spending, group savings, etc. Custom AI/ML tools can not only uncover and eliminate gender bias in general underwriting, but can also lead to better data-driven alternative underwriting.
The second subsector is made up of rural and semi-urban borrowers, where risk assessment often requires collecting data that goes far beyond the scope of the individual borrower, such as household income trends, seasonality of inflows, etc., which is also ideal for AI-based models to learn and deploy.
Overall, the power of AI/ML tools to transform how and where credit is delivered is particularly relevant and important to India’s growth story.
This article is written by Mohit Gupta, Co-founder, IndiaP2P.
