The AI Book

CHAPTER 16Using AI for Credit Assessment in Underserved Segments

By Mihriban Ersin Tekmen¹

¹Co-Founder, , Colendi

Today over 2.5 billion people are still unbanked, meaning that 38.5% of the global population is totally deprived of banking services, and only 42% of the banked population is labelled as eligible for lending.¹ The credit scoring models used by the banking system are still based on the same dynamics as when the banks were built, way back during the industrial revolution. The technical innovations of today still depend on the models as they were when first built. This explains why, in most traditional credit scoring models, the potential borrower is required to have a sufficient amount of historical credit information available to be considered “scorable”. And, in the absence of such information, the potentially creditworthy borrower often gets denied access to credit, as the credit score cannot be generated.

The main obstacle here is that the credit scoring model can only calculate a score if the potential borrower is already in the financial system, meaning that the borrower is considered to be uncreditworthy if they are unbanked. There are many reasons for people being unbanked and being unbanked or underbanked shouldn’t mean that a potential customer gets “blacklisted” by default, even if this is how the traditional credit system continues to label them. Even with the rise of FinTech, while people can access banking services without having to be a customer of ...