Artificial intelligence (AI) has been around in some form for many decades, but in the 2020s, it has matured as a combination of faster computers, big data and improvements in predictive capabilities has created a new generation of tools. Qatar has an AI strategy, prepared by the Qatar Computing Research Institute, committed to ensuring the country is a global leader in AI technologies in areas of national interest.
This article focuses on the implications for banks and financial services, where an estimate by Autonomous Next Research in 2021 anticipated nearly $500bn of cost savings globally from AI by the middle of the decade. In this context, highly relevant is the unveiling in March this year by the Qatar Central Bank of a strategic approach to create a world-class cluster of fintech businesses. Several Qatari banks have been using advances in AI, with Commercial Bank and Qatar Islamic Bank taking a lead in its applications.
AI will become more mainstream, more advanced and more embedded. Economies will have to adapt quickly and effectively in these domains to ensure continued competitiveness and prosperity.
The latest generation of AI tools are capable of processing much more data, more accurately, than any human, and can detect patterns and inform decisions. AI is, however, only helpful if the data it uses is accurate and up-to-date. Acknowledged problems with AI are machine bias – where the parameters have been affected by the bias of the system’s designers and may be magnified with use; and ‘model-drift’, which refers to a failure to update an automated system as the context changes.
If the big weakness in artificial intelligence is bad data, the big weakness in human intelligence is poor judgement – in financial services this is often fuelled by greed.
Smart uses of AI can reduce the influence of both, enhancing decision-making and reducing costs simultaneously. Three areas where AI systems are increasingly being used are investment decisions, credit profiles, and detection of fraud or money-laundering.
Investment decisions with conventional approaches are often made using a low proportion of relevant data. Smart use of AI can improve that significantly - but the final decision is still typically by a human. What the machine can do is deliver scenarios based on projections – for example, a succession of scenarios based on different levels of interest rates – and taking into account the preferred risk appetite. Moreover, with the capability to monitor millions of points of data, an automated investment adviser can detect opportunities that even an experienced investment analyst would miss.
In assessing an individual’s or company’s creditworthiness, AI models can conduct far more rigorous analysis of spending and savings histories. A decision on a loan can be both quicker and fairer. Past behaviour is generally the best guide to future risk. A bank manager’s ‘gut feeling’ can still play a role, but it is more reliable if also informed by data and analytics.
Algorithmic trading is rapid and highly effective in spotting opportunities for financial returns, but institutions may wish to monitor the wider impacts of their investment decisions, including ESG (environmental social and governance) considerations.
Some human analysis can help deal with an acknowledged problem in AI of explainability: Sometimes the automated system rejects an application, but the system cannot explain why, which does not impress the potential customer.
In checking for illegal behaviour such as money laundering or financing of terrorist groups, AI tools can be more comprehensive and accurate than traditional approaches. For example, regulations that trigger an investigation into financial transactions above a certain amount have prompted criminal gangs to use thousands of transactions below the threshold to prevent detection. An advanced AI system is capable of identifying such co-ordinated activity. It would often not be feasible for an individual or even a team to track the patterns of such a large number of transfers in a sophisticated operation.
Another social consideration is the long-discussed consideration of how automation destroys jobs. Historically, each wave of technological revolutions tends to create as well as destroy jobs, but policy makers will need to monitor the situation as certain regions and professions may be more affected than others. A report by Goldman Sachs in March this year projected that up to 300mn jobs currently carried out by people could be at risk of automation.
The experience so far in banking is that the best approaches involve a blend of human and artificial intelligence. The biggest mistakes come from failing to acknowledge their respective weaknesses. Effectively harnessing the best of both can lead to superior, even transformational, outcomes for customers, businesses and society.
The author is a Qatari banker, with many years of experience in the banking sector in senior positions.