Sandeep Anandampillai, Co-Founder & CPO, Crediwatch feels that RBI’s directives for risk management coupled with cutting-edge technology could provide the silver lining for banks plagued by bad accounts and loan fraud
As per the Reserve Bank of India’s (RBI) Financial Stability Report of July 2020, the gross NPA ratio of the scheduled commercial banks can grow significantly from 8.5% in March 2020 to 12.5% in March 2021.
Though Non Performing Assets (NPAs) have been written off by banks time and again, they continue to top the list of problems faced by financial institutions. Rising NPAs have also become a growing concern for India’s economy in recent years, threatening legitimate economic output, productivity, and jobs. Going forward, the key challenge for Indian Banks and NBFCs is to expand their credit portfolio while managing NPAs and retaining profitability. The problem of increasing NPAs in the country has put the banking industry and the lending market under immense pressure. Here, the RBI’s stringent policies have brought in some respite to the lending community. But the real game-changer will be the massive shifts in the tools of the trade, propelled by technology-led innovation.
The Tech Intervention
In 2017, the Reserve Bank of India (RBI) issued Master Directions for compliance to private and public banks and some financial institutions for monitoring their respective portfolios using the Early Warning Signals (EWS) framework. RBI and the Department of Financial Services (DFS) have defined a comprehensive framework, listing 42 and 83 signals, respectively, to be generated based on the data obtained from these sources. It is now mandatory for lenders to automate monitoring of the borrower’s health in real-time.
The emphasis from government authorities to share regulatory, tax and other relevant data vaults to the public has made the implementation of EWS possible in present-day banking. Besides, a rapidly transforming technology landscape has enabled ML-driven technology to sift through copious amounts of structured and unstructured data. Being on-alert round-the-clock allows machine learning algorithms to identify signs of distress on borrowers and classify them as Red Flagged Accounts (RFA).
How AI- and ML-powered technologies help
Most Indian business entities have scattered data points. Taken at scale, it amounts to more than 60 million MSMEs that have a negligible digital footprint. Monitoring them using traditional data points will not be sufficient to assess their risk factors. A holistic view of the borrower’s information from previously untapped sources or alternate data repositories becomes necessary. This data ranges from bank cash flow analysis, credit card usage patterns to non-banking related inputs such as utility bills, mobile or Wi-Fi bills, etc. of a business entity.
Although such data sources may seem irrelevant at first glance, they provide deep insights into responsible behavior. This is in the same way that good academic credentials indicate an employee’s potential. Hence, automated EWS takes into account all these data points while eliminating the threat of bias, errors, and high costs of manually estimating the creditworthiness of long-tail MSMEs.
- The systems pick leading signals of distress 12-18 months before the threat materializes.
- They are useful for real-time fraud identification, improved compliance, prediction of fraud and pre-emptive measures, and leveraging dynamic datasets.
- AI and ML can also play a significant role in NPA resolution. The recent RBI guidelines on NPA resolution focus on the need for a ‘resolution plan’ for accounts in default. Lenders can use ML tools to develop more meaningful and specific resolution plans that would have a higher chance of success by learning from patterns in resolution based on historic strategies, customer segments, product types, and other factors. Collection efficiencies can also be streamlined by providing insights to collection channels on successful strategies.
- AI/ML algorithms are capable of detecting and preventing suspicious or malicious transactions, enabling a shift from detective to preventive credit risk monitoring.
- Natural Language Processing (NLP) models process unstructured data such as social media, news, and alternate data for sentiment analysis.
- As algorithms improve, ML-driven EWS suppresses false-positives by learning over time and setting thresholds at optimal levels, thereby issuing more accurate results.
Prolific data, automation, and process augmentation are expected to bring transparency in auditing. This will also reduce the fears around accountability in the minds of bank managers later in the life-cycle of the loan account. Strong policy backed with analytical technology in risk management solutions will be the frontrunner in bringing the NPA woes down and boosting growth in the economy.