A New Technology Paradigm: Compliance Automation and Financial Crime Detection
How far can technology go in helping financial crime and compliance teams to introduce a new wave of effectiveness and efficiency on their mission to identify illicit activity in our financial systems?
This was a core theme that developed as part of a recent ACAMS webinar that included panelists from Commerzbank, Nasdaq and an independent industry advisor. They discussed the new face of the compliance officer and the impact that new technologies, including AI, are having across all stages of compliance, from onboarding and KYC to transaction monitoring and alert investigations.
The webinar delivered broad insights into the opportunities available, but AI technology and machine learning was a theme that ran throughout all discussions as being at the core of possibilities. The panel regularly agreed on the foundations needed to help this technology succeed, and reference was regularly made to the critical role that human teams play in its success.
The panel also recognised the market challenges that are influencing this paradigm shift. Effectiveness in identifying genuinely suspicious activity has been materially impacted by changes including rapid digitization, fragmentation of payments systems and initiatives like open banking. More recently, the impact of Covid-19 highlighted the heavy reliance on relatively immobile large teams of surveillance and investigations teams whilst criminals continue to shift with agility to find gaps to exploit our systems.
Real time, intelligent detection and investigation across all parts of compliance is the direction the panel recognised as being necessary to achieve a paradigm shift. Domains such as Transaction Monitoring can no longer rely on individual post-transaction analysis and need efficient solutions to rapidly assess and judge broader contextual data, including account customer behaviour and entity resolution.
What can technology hope to achieve?
The promises behind Artificial Intelligence are becoming real, and they underpin most of the exciting opportunities that will move us towards hyper-automation. What is increasingly highlighted in discussions like this is the role machine learning plays in advancing beyond process automation up the value chain to more complex cognitive decisioning automation.
Until recently, regulators were reluctant to embrace such advanced AI technology that is in the main pioneered by agile and disruptive startups. The viability and stability risks associated with such development have begun to be eradicated by innovation support programmes from several global regulatory bodies that provide the frameworks that banks themselves need to confidently incubate such advancements.
AI has moved on from effectively delivering simple pattern recognition and task automation to now being capable of applying human analytical traits to make complex, high-risk decisions. The challenge and limitation with such technology to date have often been the inability of AI to provide audit trails and explain itself in relation to high risk decisioning. This has likewise been a perpetual problem for human FC&C teams due simply to the nature of human decision-making being individualistically inconsistent and prone to subjectivity or emotions. This is changing through AI like Nasdaq Automated Investigator for AML that can both learn from human traits before judging risk and fully explaining decisions in human-friendly terms using those same traits.
The panelists were realistic in their assessment that the application of deeper machine learning may not necessarily result in more immediately effective financial crime detection, but there was a certainty as to the transformational speed and consistency it brings. The possible impact on existing challenges such as the management of false-positive alert volumes is also widely recognised. False positives provide huge value in helping us to continuously refine models through understanding the risks we are (or are not) looking for, but continuous optimisation of models will likely reduce the volume of repetitive false positives.
The panel discussed the viewpoint that technology has somewhat plateaued in its ability to achieve the next stage of transformation in banks. Although many institutions have deployed technology extensively across compliance, we face a situation in which the cost of both team resources and the technology itself remains incredibly high whilst the effectiveness in identifying financial crime, such as money laundering remains low.
Banks are seeking to advance both the effectiveness and efficiency of AI technology, but it is recognised that the two elements are not mutually exclusive. Their cause and effect rely heavily on one another, and the panel believed that we are on the cusp of a shift into technological maturity in which significant change will introduce further effective and valuable automation. This is not, however, about ‘automating everything.’ Hyper automation is the process of automating everything we can where incremental value exists. More importantly, it is about augmenting human teams and providing a supercharged capability focused on making better and higher quality decisions to deliver significant accuracy, consistency and efficiency benefits.
What are the enablers to technology succeeding?
The panel provided detailed insights into various aspects of current and desired compliance operations processes, but a few core themes emerged throughout that were highlighted as critical to the success of AI technology across everything from KYC and Sanctions to CDD and Alert Investigation.
Data
Data is everything. There are myriad challenges relating to everything from quality and organisation to privacy and real-time availability, but data is critical to the success of technology that relies so heavily on it if we are to effectively train and optimise higher-order cognitive machine capabilities.
The challenge to ensure that constant testing, data quality and governance is in place has become a major undertaking for any large financial institution, and banks must be ready and confident that we can consistently reference multiple data sources in an effective and relevant way. That is not limited to just internal banking data, which itself is beset by siloes, but also valuable external data sources such as the recent FinCen files, Laundramat insights or public data sources like Companies House.
To enable the desired real time quicker reactions, we must implicitly understand the data that we bring in and have confidence that technology can deal with the large volumes of data available to ingest.
Think like a human
The panel introduced one section of the webinar with the title ‘think like an investigator’ and agreed on its vital role in maximising the potential of cutting-edge technology. Whether solutions are being deployed to authenticate an individual during a simple onboarding or to establish a transaction source during a more complex investigation, technology needs to replicate the way that the best human experts think to search across the right parameters most effectively. By achieving this, a machine can then deliver that expert standard consistently and at a concurrent high-volume scale 24/7 whilst continuously learning across the compliance spectrum.
Technology understanding
Significant investment is being made by banks on technology in relation to overall compliance investment. They are also prioritising the creation and training of FCC subject matter experts who can benefit from the augmentation that advanced AI delivers. It is essential, however, that those experts understand technology, data and machine analytics. How they work and what they can do are critical to the scoping, training, implementation, management, governance and optimisation of the machine.
Likewise, if compliance teams are to be augmented to deliver better quality decisions, then surveillance and risk investigation experts need to know how to use it alongside the more complex tasks they deliver. Finally, they must work cohesively across data science, risk and audit teams to get the right momentum and collective output that eradicates siloes from processes.
Tipping point
The sense from this session was that we are reaching a tipping point of technology maturity. A point at which the scale and speed of machines can be powered by human intelligence to create more valuable and effective automation of high risk decisioning.
Our first-hand experience with Nasdaq Automated Investigator for AML has established the benefits that this advancement in AI can bring to effectiveness and efficiency in a critical compliance area of investigations.
We also appreciate first-hand the data and technology learning curve that is required to ensure success. There are complex challenges to overcome, but the prizes are significant, and we collectively look forward to these panel discussions around the technology paradigm becoming a wide-reaching reality in the not too distant future.
Other Topics
Artificial Intelligence
MarketInsite
Nasdaq
Nasdaq’s Marketinsite offers actionable insights on a variety of market-moving topics. Learn from our thought leaders who are driving the capital markets of tomorrow.
Read MarketInsite's Bio