AI and machine learning are reimagining risk management in FNB

In recent years, FNB Risk has embarked on a data-driven strategic initiative aimed at enhancing and connecting insights across risk business units while driving efficiency, eliminating redundant activities, keeping up with complex, evolving, and emerging (Climate, cyber, financial crimes, model, etc.) risks, and freeing up risk and other professionals to do what they do best.

Many of the skills driving this strategy are enabled (and augmented) by artificial intelligence (AI) and machine learning (ML), which have become an unavoidable component of the FNB risk function and its modernization. Why? Because ML can learn from experience and AI is capable of processing enormous volumes of data while recognising patterns and identifying abnormalities that humans cannot.

Similarly, ML may assist in detecting fraud, combating identity theft, and combating other forms of wrongdoing by identifying behavioural anomalies in big databases. Predictions may also benefit from ML and AI since everything learnt from prior cases can be applied to future ones, with a decreasing proportion of false positives.

AI and machine learning can also assist in combating the security concerns posed by an increasing number of internet-connected devices mixed with globalisation – a combination that opens the door to an increasing number of assaults from worldwide enemies.

AI and ML enable more than merely cost savings, though. They are enabling the risk function to build more desirable products for its customers and improve customer experience by, for example, streamlining the onboarding process and flagging problematic documents through AI enabled ID verification models at submission rather than after the fact. They also empower the Bank to improve our products and services over time by taking into account the ways customers use them, and any problems they experience along the way.

These new technologies also enable FNB to conduct real-time monitoring, as well as deep insights and analytics that were previously unachievable. At the same time, they alleviate the burden of time-consuming manual evaluations. On average, AI saves 70% of analysts’ time by providing a forensic summary ready for a human analyst to evaluate. What used to take hours may now be performed in as little as eight seconds.

This frees up the risk team to conduct deeper dives and identify root causes, while also positioning us to identify emerging risks. We’re also able to meet regulatory requirements and make forensic due diligence decisions faster, more accurately, and more efficiently by leveraging our internally developed AI system. This AI system has been scaled to the rest of Africa, where it’s being used for fraud investigation, suspicious transaction reporting, and even to enable the risk advisory space.

AI is also being used for business intelligence by enabling the automation of risk event insights and risk decisioning. Risks worthy of escalation can be identified more accurately and rapidly, and business unit-specific risks can be identified, while the learnings from them can be transferred to other business units.

At the same time, these sorts of risk models are proving useful outside of customer interactions. For example, AI and algorithmic risk assessment can be invaluable in the realm of climate risk assessment where myriad variables need to be considered in tandem with one another. The forecasts will enable new strategies and business models that can account for climate risk, something that’s previously been arduous or impossible.

The FNB Risk data literacy programme, which aims to equip every member of the organisation with the necessary skills to turn data into actionable insights, has not only enabled a growing number of our colleagues to harness AI, ML, and other emerging technologies to deal with existing challenges, but it is also equipping them to deal with new ones. The curriculum has been ongoing for 18 months and was meant to empower the whole risk workforce, regardless of job or prior expertise with data analysis, to make data-informed choices.

The above-mentioned results are a direct outcome of the risk data strategy’s strategic objectives, which were established three years ago. Among these were leveraging data to improve risk management practises and proactively detect risks that exceeded risk appetite and tolerance levels, automating manual risk management procedures, generating effective data risk aggregation and controls, and allowing end-to-end data governance.

These same methods will allow the next frontier of risk data asset generation, management, and data-driven risk decisioning over the next 18 months.  They’ll enable us to build further and scale our AI capabilities, increase AI and data analytics literacy across the group (especially for risk), drive collaborative engagement across FNB, and position FNB to continue reimagining risk in the future, and solve for as-yet unforeseen risks.