Capital markets professionals are convinced AI is going to transform credit decision making but some are worried about the ethical and regulatory implications of such a development.
In the latest indication of just how pervasive AI is set to become in our lives, capital markets professionals say they believe the technology will produce more accurate, reliable and transparent credit decisions than human-based systems within five years.
The finding, based on a survey of over 500 capital markets executives by Intertrust, the Amsterdam-listed administrative services provider, says 14% or one in six of respondents believe that AI is already outperforming human-based systems. While certain of the growing role AI is set to play in financial services, the executives are also concerned about its regulation and its impact on rights. Over a third believe tighter legislation is required to protect borrowers’ rights, while 20% believe the use of non-traditional data has already overstepped the ethical line.
The survey, aimed at identifying the impact that disruptive technology is having on jobs and skills, notes that in recent years the data sources used in credit decision-making have become increasingly broad and non-traditional, now including social media activity, retail spending habits and even political inclinations.
The research revealed a division in the industry about the impact of using such data on the quality of decision-making. While a third (30%) of respondents believe that using a broader range of data reduces subjectivity, a fifth (18%) think AI exacerbates existing prejudices in the credit decision-making process.
Intertrust’s study also highlighted privacy concerns regarding expanded data sets. Although almost a third (31%) of respondents think that the use of non-traditional data such as that derived from personalised algorithms leads to better credit decisions than just relying on detached data, 36% believe tighter legislation is required to protect borrowers’ rights when they apply for funding and to restrict the information included in the assessment.
Cliff Pearce, Global Head of Capital Markets at Intertrust said, “The use of AI in credit decision-making has become increasingly commonplace, with the potential to make quicker, more accurate credit decisions, based on an expanded set of available data.
“A challenge in this area is that AI systems are only as good as the information programmed into them. For example, while a prospect may look like a poor risk at first sight, there may be extenuating circumstances overlooked by the system that a human would have noted. Put simply, AI underlines the contrast between the prime and more specialised non-conforming lending markets.”
Drowning in data
The potential impact of AI across the consumer lending universe, which includes mortgages, cars, credit cards and student loans, is enormous, as there is so much data generated. US-based Tavant, an IT solutions and services provider which employs AI and machine learning (ML) techniques, estimates, for example, that in a typical mortgage more than 5,000 data attributes are captured during the lending process, covering the borrower’s credit history, property, employment, income, tax and insurance information.
“That is a time-consuming and expensive process and in the case of many lenders, a very manual one. How much of this data is even relevant? How much of it is useful in predicting borrower behaviour during the application, closing, post funding and servicing stages? As such these are questions that are perfect applications for machine language to answer,” says Tavant in a white paper published last year.
It adds: “Given that many financial institutions have many years of this data stored or accessible makes it perfect to train using ML algorithms to devour and tell us what is useful, predictive and how. Most companies infer only a small fraction of the vast potential of what is helpful from this data. This is primarily because they are mostly using traditional reporting techniques to analyse it, if at all.”
