iscover
Access to more meaningful data, combined with the application of machine learning (or AI more broadly), will improve the credit underwriting process and result in more accurate predictions of defaults and potential losses, ratings agency Standard & Poor’s has claimed.
However, S&P’s recent report, titled Australian Structured Finance: Credit Analysis in a Digital Ecosystem, notes that the use of past credit-cycle data is needed for lenders to make accurate credit analyses.
As digital lending is relatively new and “collateral performance has not yet been observed throughout economic cycles”, it could take some time for digital lenders to realise the full benefits of AI, despite them being “the new data pioneers in incorporating ‘alternate data’ into credit decision-making”, according to the ratings agency.
“According to the US Treasury Department, ‘Machine-learning models… would generally suffer from the absence of past credit-cycle data to train the model’,” the S&P report states.
However, the mandatory comprehensive credit reporting (CCR) and open banking regimes could change that, allowing online lenders and other fintechs to compete more fiercely with the incumbents.
“Fintech companies are using their sophisticated platforms to expand customer networks and develop detailed borrower profiles based on customers’ data. This gives them a clear advantage over the financial services incumbents, which often have cumbersome legacy systems, and enables them to target their market share expansion and be more agile in their scope and coverage,” the S&P report states.
“With open banking and [CCR] on Australia’s doorstep, the penetration of fintech companies will only increase as they gain access to more of borrowers’ financial data.”
Applying AI to loan servicing could also give rise to “proactive” loan servicing, the ratings agency said.
Currently, loan servicing is reactive in that systems currently only track payments and generate alerts when borrowers have failed to make payments on time.
With access to more data points, machine learning-powered systems could predict changes in the borrower’s spending patterns and allow lenders to “step in early and customise a solution to borrowers’ financial situations before they miss a payment”, S&P said.
AI could also make bespoke loan products (such as non-conforming loans and SME loans), which are considered “more labour-intensive”, more scalable.
“Credit decision-making processes often rely on staff making additional checks to complete borrower credit assessments, depending on the complexity of the credit and the bespoke nature of a loan or borrower. Access to more data will reduce lenders’ reliance on more manual checks,” S&P’s report states.
“This will make operations more scalable and reduce the risk of deterioration in lending standards during periods of strong growth, when headcount might not keep pace with lending growth.”
Further, AI technology could also make lending to small businesses, a market that is well-known to have difficulty in accessing credit, more attractive for lenders.
“The deployment of sophisticated technology platforms and access to more customer data will enable lenders to lower the cost of credit-risk assessments and allow more comprehensive risk analysis for pricing, thereby improving borrowers’ access to finance,” S&P’s report states.
A heavy reliance on the use of AI to make credit decisions, however, could make it difficult for regulators to audit and interrogate decision-making processes in the future.
“This risk needs to be actively managed to ensure digital platforms driven by AI and machine learning don’t become a new type of ‘black box’,” the report states.
[Related: CCR could be a ‘gold standard’ for credit risk]
ISCOVER 
