As the number of potential borrowers continues to rapidly grow, loan companies and banks are having a bad time trying to figure out how likely their customers are to pay back. Probably, getting information on clients’ creditworthiness is the greatest challenge for most financial companies, and it especially concerns those clients who don’t have any credit history yet.
There is no denying that the alternative lending business has become one of the most influential financial branches both in the USA and Europe. Debt is a huge business of our days that needs a lot of resources. In such a challenging situation, any means that can improve productivity and reduce the risk of mistake while performing financial activities are warmly welcomed. This is actually how Artificial Intelligence became the redemption for loan providers.
Fortunately for lenders, AI successfully deals with this task by following the borrowers’ digital footprint. For example, some applications for digital lending collect and analyze an individual’s web browsing history (upon receiving their personal agreement on the use of this information). In some countries such as China and Africa, they may also look through their social network profiles, geolocation data, and the messages sent to friends and family, counting the number of punctuation mistakes. The collected information helps loan providers make the right decision on their clients’ creditworthiness and avoid long loan processes.
When AI Overfits
Unfortunately, there is the other side of the coin. There’s a theory which states that people who pay for their gas inside the petrol station, not at the pump, are usually smokers. And that is the group whose creditworthiness is estimated to be low. But what if this poor guy simply wanted to buy a Snickers?
This example shows that if a lender leaves without checking the information carefully gathered by AI software, they may easily end up with making bad mistakes and misinterpretations. Artificial Intelligence in the financial sector may significantly reduce costs, efforts, and further financial complications, but there are hidden social costs such as the above. A robust analysis, design, implementation and feedback framework is necessary to meaningfully counter AI bias.
Other Use Cases for AI in Finances
Of course, there are also enough examples of how AI helps to improve customer experience in the financial sector. Some startups use AI software to help clients find the company that is the best at providing them with the required service. They juxtapose the clients’ requirements with the companies’ services finding perfect matches. Even though this technology reminds us of how dating apps work, such applications can drastically save time for both parties and help borrowers pay faster.
AI can also be used for streamlining finances. AI helps banks and alternative lending companies in automating some of their working processes such as basic customer service, contract management, or transactions monitoring.
A good example is Upstart, the pet project of two former Google employees. The startup was originally aimed to help young people lacking the credit history, to get a loan or any other kind of financial support. For this purpose, the company uses the clients’ educational background and experience, taking into account things such as their attained degrees and school/university attendance.
However, such approach to lending may end up being a little snobbish: it can simply overlook large groups of population who can’t afford higher education. As a result of insufficient educational background, these people can become deprived of the opportunity to get their loan.
Nonetheless, one of the main goals of the company was automating as many of its operating procedures as possible. By 2018, more than 60% of all their loans had been fully automated with more to come.
We cannot automate fairness and opportunity, yet
The implementation of machine learning in providing loans by checking the digital footprint of people may lead to ethical and legal disputes. Even today some people state that the use of AI in the financial sector encouraged inequality in the number of loans provided to the black and white population of the USA. They believe that AI continues the bias against minorities and make the black people “underbanked.”
Both lending companies and banks should remember that the quality of work done these days with the help of machine learning methods highly depends on people—both employees who use the software and AI developers who create and fine-tune it. So we should see AI in loan management as a useful tool—but not as a replacement for humans.
Darya Shmat is a business development representative at Iflexion, where Darya expertly applies 10+ years of practical experience to help banking and financial industry clients find the right development or QA solution.