Like in any other industries, lending is also betting big on Artificial Intelligence (AI). However, if we observe most seminars, thought-leadership discussions, or write-ups, the same few topics are discussed repeatedly:
- Alternate Data for Underwriting – Using all possible users’ data such as utility bill payments to social medial actions to e-commerce purchases to derive the credit worthiness of the individuals
- Risk Based Pricing – India is a credit hungry country. Many individuals who want to break the glass ceiling are denied lending products due to standard lending norms. A differentiated pricing for additional risks by lenders will help to bridge the gap between demand and supply of loans. AI companies are working hard to identify the higher risk in customer profiling and recommending additional pricing for the same.
- Funnel Optimization – Digital loans thrive on conversions. Higher the conversion rate, lower the sourcing cost is. And hence, the industry is infatuated about optimizing the funnel conversion.
- Fraud Detection – Identifying the document and mismatch related frauds. Though the opportunities are large, we are still scratching the surface. Now regulators are questioning the authenticity of these service providers too.
- Document reading and linkage – OCR is our new best friend. Our quest is to train AI to read a document and match the information with other datasets.
This has been the discussion agenda for a decade now. And most of it is already available to some extent. Of course, finetuning can be done. But they cannot serve as game changers any longer.
Then, what should AI solve now?
1. Fundraising
Lending needs availability of capital with lenders. With savings accounts becoming less attractive for individuals, even Banks are struggling to raise funds for lending. NBFCs are even facing harder times.
Equity raise is equally competitive. Every investor looks at compliance with regulators, NPAs, book concentration and operating cost. The math’s often does not stack up for investors when it comes to funding lending institutions.
2. Credit Assessment
Various challenges that lenders grapple with to assess borrowers are –
- No comprehensive database for PEP (politically exposed person) related search. Today, the world is connected like never before. A simple car loan borrower may have a connection with a local politician. It is not necessary to limit PEP search with nomination filing only.
- Bureaus speak different languages. The four bureaus never share the same information for any borrower. Either the count of inquiries or DPD (no of days, account is in overdue) show mismatch.
- Defaulting first EMI after the loan is taken is treated at par with subsequent EMI bounce. First EMI bounce is a clear intent issue and brings the different level of seriousness in the lending community.
3. Compliance Costs
Not that I disagree with RBI’s (Reserve Bank of India) view of protecting borrowers’ interest. However, there must be balance. Deleting data or obtaining / refreshing information time and again does not sit well with smaller players. They have to either rely on a tool (which does not make sense considering their volume) or make it human driven (which is against prone to errors).
4. Early Warning Signals to Exit Plan
AI should be leveraged for various EWS flags and propose the strategy for lenders.
AI should also propose the probability of default for borrowers based on these emerging red flags.
These predictive insights could automate parts of the workflow and even drive early, meaningful borrower engagement.
5. Collection & Legal Cost
One delinquent borrower and multiple legal cases. And each lender pursues their legal action independently. If the underlying charges are the same, why not have this cost combined? While the fact is, lenders rarely win any of these cases against borrowers. Getting any order against borrowers due to cheque bounce is still a distant dream for lenders.
Apart from legal cost, there is a recovery cost which is multiplied against the multiple lenders as all of them are trying to recover from the same borrowers.
Conclusion
AI companies need to solve beyond the obvious. These are not new challenges but ignored ones. It’s time that we pick up a genuine problem to solve. Current evolution can continue to finetune the existing models. But higher bandwidth is to be used for solving grassroot problems.
Also Read- https://www.businessoutreach.in/nach-mandate-issues-for-lenders/
Authored by
Swapnil Singh
SVP – Risk Management & Operations, CredAble
Views here are personal
