Nic Swingler, Head: Financial Crime Compliance

The potential benefits of using artificial intelligence and data analytics in making banking safer are a constant theme. In reality, however, it is hard to develop the systems and processes required to turn potential benefit into actual benefit.
Absa’s separation from Barclays — a three-year, multibillion rand programme across 12 countries — presented a rare opportunity to do exactly this.

Financial crime is nothing new: It’s been around since the day people stopped bartering and started using gold as a payment system. However, the way in which financial crimes are committed today is much more sophisticated than shaving some gold off a coin to cheat a seller.

The rules sound simple, but in practice it requires the assessment of many hundreds of thousand transactions or customer activities each day.

Money laundering, tax evasion, embezzlement, forgery, counterfeiting, hacking, phishing, vishing and identity theft are just some of the crimes that banks and authorities fight every day and banks’ systems must evolve constantly to remain a step ahead in ensuring that activities are detected and reported to the appropriate authorities.

In fact, the Financial Intelligence Centre Act (Fica) requires that “intermediaries in the financial system must know with whom they are doing business”. Simply put, banks need to know who their customers are, as well as their banking patterns and behaviours.

Strange activity

This goes well beyond just checking a face against an ID document; it also means ensuring that systems can detect “strange” or unusual account activity, such as large sums of money moving around rapidly without an obvious reason.

The rules sound simple, but in practice it requires the assessment of many hundreds of thousand transactions or customer activities each day, just to be compliant.

The traditional way in which to do this is through automated systems, with a set of behavioural rules that flags suspicious or unusual transactions or activities. These flagged transactions are referred to analysts who review the activity more closely.

However, “false positives” are often raised in cases where transactions are flagged but turn out not to be suspicious or unusual, which wastes time and valuable resources. On the other hand, the behavioural rules need to be comprehensive and complete to guard against the risk of not detecting all suspicious or unusual activities.

Absa’s separation from Barclays — which was substantially completed in June after three years — gave us the chance to consider alternative solutions. We implemented a solution which enabled the use of more sophisticated behavioural rules, AI and network analysis, and which could be customised to ensure it was relevant to each of our operations. There were some significant execution challenges but this solution was implemented on time and within budget.

Importantly, we were able to move from just being compliant to becoming more effective at managing financial crime risk. This goes beyond transaction monitoring and knowing our customers, and calls for proactive risk management, data and network analysis as well as investigative and intelligence gathering capabilities. It’s more than just a system or platform; this risk management approach also requires specific human skill sets, which need to work in tandem with the technology solution.

Absa’s separation programme presented us with a rare and valuable opportunity to bring about a step-change in the quality of our financial crime platform and systems.

The solution we implemented allows for customisation of rules, an improved user interface and a more sophisticated ability to design behavioural profiles. It also enabled Absa to add on plug-ins based on data analytics, derived from behaviour, which refined the system’s ability to flag potential financial crime activities.

With the deployment of artificial intelligence tools, we were able to significantly reduce the time spent in assessing alerts and rather re-invest this time in high value risk management activities.

Continuous learning

The system also takes into account customer behaviour, which further improves efficiency and it produces better-quality alerts using a sophisticated behavioural algorithm that continues learning through AI.

Absa’s separation programme presented us with a rare and valuable opportunity to bring about a step-change in the quality of our financial crime platform and systems, which ultimately benefits our stakeholders and customers. And while it doesn’t mean that all risks have been eliminated, our ability to surface and analyse issues has meant that we are better able to manage and mitigate the risks and threats, helping to make Absa a more efficient and a safer bank.

https://techcentral.co.za/how-absa-is-using-ai-to-fight-financial-crime/101647