The reliance of the financial industry on both customer volume and complex decision-making makes it a prime candidate for embracing this artificial intelligence (AI) and machine learning (ML). Specifically retail banking, which relies on earning a small interest rate spread across a vast number of clients, automation and optimisation in business operations would provide a significant competitive advantage.
On the front-office side, AI has already made huge inroads into providing automated services to customers, in areas such as identity verification, fraud prevention and customer assistance. This involves technology such as chatbots, voice verification and other automated banking activities.
This not only has the benefit of reducing costs involved in servicing a growing client base, but as confirmed by an Accenture survey, caters to a growing client demand – 71% of their sample would support a move to completely online banking. This automation reduces client waiting times for support, identification and opening new accounts, increasing the ease of doing business and hence creating a significant growth opportunity for financial services.
Beyond the client-facing portions of retail banking, AI has significant potential to be applied in processing client applications and in analytics – both identifying opportunities and preventing loan impairments.
Machine learning involves models that improve through use – in effect ‘learning through doing’ – this presents an opportunity for a more sophisticated analysis of credit risk, when deciding on lending. Updating current credit extension criteria to include deep learning models, could allow banks to identify both risks and opportunities missed by current methodology.
AI in retail banking – data modelling
Large tech companies naturally collect vast quantities of data as a by-product of their business, allowing them to feed this into AI modelling for credit risk. Recently, Amazon partnered with Goldman Sachs’ commercial bank, Marcus, to provide loans to merchants on its platform, underwritten by Marcus; the underwriting is driven by Goldman’s credit risk model that, as part of the agreement, is fed with data collected from Amazon’s platform.
It is plausible that improvements in AI will allow for reduced collateral requirements, due to the improvement in credit-quality assessment, presenting an opportunity for banks to extend lending to households and firms that would traditionally be rejected, driving loan book growth. Furthermore, due to its ability to detect patterns indiscernible to loan managers by analysing huge datasets, AI could also be used in identifying riskier loans that have been extended, saving the firm from potential loan impairments. Through such a process, financial institutions have the potential to grow their loan book without compromising on credit quality.
This process is both automated and adaptable to changing consumer trends. The ML model is dynamic, allowing it to respond to changes in the underlying data in live time, completely autonomously. This would not only be valuable from a risk management perspective in preventing potential default, but it would also allow a bank with such technology to strategically target its products towards certain businesses, industries, locations, etc. – allowing it to more clearly realise where the demand for credit lies.
 Accenture Financial Services, 2017 Global Distribution & Marketing Consumer Study: Financial Services Report
 Congressional Research Service (2022), “Big Tech in Financial Services”
 BIS (2020), “Data vs Collateral”, Working Paper No. 881in