The rise of Artificial Intelligence in consumer lending

More and more lending institutions around the world are moving to either fully automated or partially automated systems when assessing consumers for home loans, personal loans, credit cards pay-day loans and other types of consumer finance.

Part of rise in Artificial Intelligence (AI) algorithms assisting in finance comes from a need to increase efficiencies and processing times and part of the rise is also to do with the increased need for regulation technology to remain compliant under a changing regulatory landscape world-wide.

AI and banking
AI and banking: Lending institutions around the world are moving to either fully automated or partially automated systems.

One of the leaders in AI technology to assess credit applications is Advantage Data who have being building credit models in Australia in partnership with a number of online loans providers. The model considers applications bank statements, loan application and credit history.
However, the success of these models is often in the detail of what’s been analysed and Advantage Data use advanced text analysis techniques to process the free text that loan applications enter. Much can be gleaned including personality traits and the likelihood of fake loan reasons which all effect the probability of a customer paying back a loan.

Other lenders take this idea further and examine customers customer’s social media profiles and the data they maintain when assessing loans. There is so much useful information within your social media profile that can be used to determine the likelihood of you obtain finance. Most people have a real issue when presented with this fact, but social media profiling can often assist applicants as well. All manner of data is analysed, even including the number of friends you have, past job history etc.

The Cambridge Analytica affair will put the spotlight on social media data, but any information you put in the public domain is often fair game when it comes to third parties using that data.
Where you live can also affect your chance of getting a loan. Most developed countries including the US have detailed demographic information down to street level. Often things like where you stand vs. your suburb’s annual income is looked at. Also, if you have moved, some models consider whether you have moved to a better suburb or a worse suburb, as a proxy for how successful you have been with your finances if you have been renting. If you have purchased a house in a lessor suburb that’s usually fully accounted for in the modelling.

AI and big data can assess much more information than a human can, and this creates efficiencies for lenders. Often this technology is also in conjunction with manual supervision. For example, the AI technology might give a rating for a customer and then customers over a threshold rating are passed to manual loan assessor.

Regulators are grappling with this emerging technology as well. This has spawned the concept of Explainable AI where AI algorithms that were previously black box (even the designers don’t know what variables impact the predictions in what way) now need to be interpreted for the benefit or regulators. Sometimes building models that can be interpreted results in a loss of accuracy compared to the old school black box models, which is a compliance cost. Another area of data science that’s emerging is building models specifically to interpret other models to justify decision making.