Decoding The World Of Fintech: Role Of AI In Credit Line Assessment
Ask anyone who knows a thing or two about technology, and they will tell you that AI has finally started paying the dividends in the vast battlefield of finance. According to Eric VonDohlen (VP, BI & Analytics, ICW Group), a neural network is the only artificial thing that can mimic a human brain with fantastic accuracy. While driving the point home, he goes further to classify traditional linear models as dogmatic, and as something which wants to impose structure on data rather than listen to it. Poignant words indeed. In fact, in the world of credit line assessment, a regression model powered scorecard is dogma. Even more so in the commercial lending space, ergo, B2B. All with good reason though. Lending has always been a risky business, and banks take extra steps to ensure they don’t run into losses. Even when lending to businesses.
Where do we stand today?
If you apply for a loan today, the bank will look at your credit rating. If it’s between the stated limit for that particular loan, then you will be asked to produce identification, and past bank statements, along with other necessary financial details. All of this will go into a model that will crunch all the variables, calculate the numbers, and then come up with a Yes/No decision on the loan disbursal. Businesses will also have to go through a similar process, although the documents and conditions might be different.
By crunching the variables, the model’s algorithms will look for relationships and connections that are out of the ordinary like pending loan payments, property debts, etc that can scuttle the chances of a positive decision. Powered by various types of statistical regression algorithms, the models also throw up the variable that influenced the decision. This is then communicated to the customer as the reason behind the rejection. In countries like the US, banks need to provide loan seekers with the reason which is also known as Adverse Action Reasoning (FCRA). Complete transparency with minimal chances of error. Even your credit scoring system works on the same regression principles.
So what is the problem?
There are several holes in this process. The first thing would be the fact that most logistic regression models are linear, and miss out on all the hidden relationships between variables. Simply put, it’s completely blind to your social data, your internal relationships, etc that should influence the decision. The same goes for any business loan or B2B loan as well. The underlying principles remain the same.
The first problem segues into the second issue, which is accuracy. Although it can be good with these logistic models, they can be made much better. Due to sub-optimal accuracy, a lot of creditworthy individuals and institutions can be left out of the credit fold altogether.
The third problem lies again with the data. Nearly 94% of banks don’t leverage 3rd party data such as LexisNexis which adds to the problems. Other issues lie with the models itself where they use a limited number of features and predictors which sometimes work on customer data that’s more than 5 years old.
However, artificial intelligence can change all that, and disrupt the space forever.
What business leaders in the BFSI sector should know?
With pioneering advancements made by some ingenious start-ups like episenseAI and scienaptic.ai in the field, artificial intelligence solutions are helping lending institutions make sharper underwriting decisions by leveraging variables that factor more accurately in assessing millennials. These AI hubs in the fintech space have already helping auto-lenders, alternative lending firms, and banks use machine learning algorithms to cut significant losses annually.
With certain innovations, like automated AI powered platforms by episenseAI, lending institutions can now finally adhere to FCRA and SR 11-7 regulations while using AI. No longer do machine learning algorithms operate in blackbox mode, and that allows banks to freely adopt AI in their credit line assessment processes.
This way the processes remain transparent, but get a major boost to accuracy thereby bringing more people into the credit fold. A win-win situation as the banks can now reduce their bad rates even further, increase their customer base by disbursing more loans, and improve net profits. A far cry from today’s traditional processes that hampers credit availability to millions in the US where nearly 24% of the population have no FICO score or credit lines.
Adoption of machine learning by the BFSI sector
Adoption is taking place, albeit at a very slow space. However, the pace has picked up, and 2019 saw a lot of institutions like Capital One banking more on leveraging their machine learning resources to develop products for credit line assessments. With powerful predictive machine learning algorithms, lending institutions are using broader datasets to help expand the ultimate goal of financial inclusion.
New research from Refinitiv is even more heartening which states that nearly 90% of BFSI firms have begun using machine learning in multiple areas as a core part of their business. While 75% of these firms are making significant investments in machine learning, nearly 62% of C-suite respondents plan to hire more data scientists as banks ready themselves to gain an upper hand in the competition.
In fact, BFSI firms like Discover, MasterCard, etc have already begun applying machine learning to reduce fraud, check kiting, and credit assessments.
The final word
Although AI has begun reaping the dividends, the pace is still too slow. With credit scoring companies like FICO coming up with unique AI-powered solutions that leverage machine learning, it’s now or never. To help make AI mainstream in banking, platforms need to be automated and flexible. Some other innovative start-ups and companies like Episense, Kreditech, and Lenddo are coming up with their own automated AI powered platforms that are highly flexible, robust, and operate in whitebox. So, if you were wondering where AI stands in credit line assessment today, then by now, you might have gotten your answer. It’s standing tall, ready to take the BFSI sector to an entirely new level as we move further into 2020.