Merging a rules-based approach with AI in finance

By KK Gupta, CEO. Facctum

Of all the technological innovation that has taken place over the last decade, one of the most exciting is the emergence of artificial intelligence (AI), which is revolutionising practices across the banking sector. For understandable reasons, leaders in the sector are keen to leverage the power of AI. The technology has, for example, considerable power to automate manual workloads and boost efficiency. AI can also be used in regtech systems, monitoring transactions to spot outliners, thereby boosting firms’ anti-money laundering procedures and preventing fraudulent activity from taking place.

Most financial institutions are keen to ensure that they understand AI and that they do not get left behind – not least because doing so could put them at a competitive disadvantage. As a result, there has been a rush to adopt systems, with at least 85% of financial institutions using AI in some form. In terms of handling data, there has been a resulting shift away from traditional “rules-based” systems, which are programmed to follow prescribed criteria when analysing data, and towards the use of AI on its own.

This trend is showing no sign of slowing down. Surveys have shown that banks and insurance companies expect an 86% increase in AI-related investments by 2025. But while investment in innovative technology can yield significant benefits, there is a risk that this rush to adopt AI is leaving a number of gaps.

Potential risks

Because much of the current hype on the market today is directed at AI, many businesses might forget the fact that a rules-based system still commands very significant benefits. Rules-based technology remains, after all, an innovative form of data science in which there is considerable investment and innovation taking place.

Perhaps most importantly, rules-based systems are much easier to understand and explain to a regulator. Should an institution be asked to explain why their systems made a certain decision, they can easily demonstrate the criteria which was used. Undertaking the same task with an AI system is much more difficult, because they make decisions without referencing predetermined and transparent criteria. Explaining these decisions, should the system fail or be perceived as unfair, is therefore tricky, and this can land firms in trouble with regulators.

Indeed, AI has been criticised for worsening discrimination against minority groups trying to apply for loans. Google looked into whether AI could help firms decide whom to lend money to, but halted the project because it was deemed too ethically risky. Recognising the potential problems associated with AI, financial regulators in the UK have warned banks that they can only leverage the technology if they can implement the necessary safeguarding measures to guarantee such biases are not perpetuated. They must also be able to explain how their technology is making decisions.

Partly because of these experiences, regulators are asking more and more questions about AI. How does it work? How often do you review it? How do you ensure it’s giving you accurate results? How does its effectiveness measure up compared to your risk profile? Answering such questions is straightforward if clearly defined rules are in place that stipulate how software can analyse data – rules that provide a regulatory foot-trail. But firms using AI will need to do a lot of laborious work if they are to rewind the original decisions and put it in the relevant context. In situations where the outcome is the same when using rules, compared to using AI, rules are better for this reason.

A blended approach

Of course, this is not to imply that AI does not have a wide range of use-cases in financial institutions. Rather, it is simply to suggest that a blended approach – merging rules with AI – may be preferable. For example, a strong and transparent rules-based approach can ensure very high levels of risk detection, with the power of AI then being leveraged in post-processes.

And indeed, it is by automating these post-processes that AI can add particular value. AI can automate a wide range of manual tasks, which can be timely and expensive, and therefore make an entire business process cheaper and quicker. More than a third (37%) of financial services firms have reported that their organisation has reduced operational costs because of AI adoption, with a further 34% predicting that AI will lower their cost base over the long-term.

Innovative use-cases

When it comes to risk management, increased speed and automation has the added benefit of reducing the compliance gap – the time it takes between a risk being detected and then it being validated. Combining a rules-based approach to risk detection, an approach which is fully transparent, accessible, and explainable, with the transformative power of AI has the potential to facilitate compliance and risk management strategies that are more effective and more advanced than ever before.

It is also true that AI has a large number of other use-cases, aside from compliance and risk management, that can significantly enhance the customer experience. Automation, predictive analytics, and AI can prevent customers from having to fill out long forms during onboarding processes, for example, reducing customer friction and boosting successful onboarding rates. Research shows that 32% of firms which deploy AI in this way have already reported enhanced customer service and satisfaction.

For banks themselves, AI can also help them offer the optimal products and services to customers much more quickly and effectively. In turn, this allows banks to prevent potential customers from getting stuck in lengthy and drawn-out KYC and onboarding processes. Instead, accounts can be grown from the very start, with customers immediately starting to make use of all the services a bank might offer. In other words, appropriate deployment of AI can create a much smoother customer experience, while boosting productivity for banks as well.

For all these reasons, AI is clearly one of the most innovative developments in recent decades – and the innovation is set to continue. But for all of its undoubted benefits, it is not a silver bullet. There will always be areas where strong rules are still needed, where the human capital of well-trained experts is still highly valuable, and where business processes and corporate cultures still remain crucially important. If those pieces are not in place, the technology simply will not reach its full potential and the inevitable result is that AI will not be as transformative as it otherwise could be.

Combining AI with another form of innovative data science, rules-based technology, is the ideal way for financial institutions to take full advantage of new technological trends without exposing themselves to unnecessary regulatory pressure.