‘Be watchful’: RBI Governor Das on use of AI and algos in consumer loan landscape

The Reserve Bank of India (RBI) is urging banks, NBFCs, and fintech firms to recalibrate their pre-set algorithms applying artificial intelligence (AI) and, increasingly, utilising machine learning tools periodically. This should be done in response to the evolving dynamics of the financial ecosystem and new information emerging about sectors, consumer segments and global headwinds.

“Banks and NBFCs need to be careful when relying solely on pre-set algorithms as assumptions based on which the models are operated. These models should be robust, tested, and re-tested periodically,” said RBI Governor Shaktikanta Das.

Das was speaking at the FIBAC 2023 Conference, organised jointly by FICCI and IBA in Mumbai.

“They may need to be calibrated and re-calibrated from time to time based on the changing contours of the financial ecosystem and fresh information. It is necessary to be watchful of any undue risk build-up in the system due to information gaps in these models, which may cause dilution of underwriting standards,” added the governor.

Fintechs, once touted as disruptors, now collaborate extensively with established banks for lead generation and co-lending. This new and fast emerging relationship benefits both parties, providing banks with ready customers at no operating cost and Fintechs with a broader platform and lending capacity.

However, this partnership has spawned innovative products and services, along with new business models. Banks and NBFCs, for instance, can now fund ‘new to credit’ customers by analysing their buying behaviour trends, with fintechs leveraging hundreds of data points for creditworthiness assessment.

In addition, cash flow lending to small entrepreneurs in smaller geographies has emerged. However, machine learning or AI-based models have inherent limitations, as they rely heavily on historical data and producing outputs based on the input fed into the system. It is not known how the new data or developments are fed into the system and at what intervals.

While the RBI Governor has highlighted concerns about model-based lending through analytics, other issues, such as ‘exclusions’ or biases introduced by machine decision-making, also warrant attention.

Also Read: Tata Technologies IPO oversubscribed on Day 1; here’s how stocks from Tata group performed since 2004