A banker wants to induct a new client, Corporate X, for the bank. His colleague in Risk taps a few keys on the Risk Portal…and lo! All the relevant details are immediately flashed…Corporate X’s banking history, credibility, the bank’s funding appetite and recommendations. The portal even calls out red flags such as a likely management shake-up (deciphered from social media); potential litigation and even industry headwinds.
Seems like a scene out of a Sci-fi movie? Welcome to the future, for this could be the likely scenario of the BFSI sector, thanks to Artificial Intelligence (AI).
AI is the ability of a computer or robot to perform a task that requires human intelligence – assimilating and processing data, making decisions, or challenging choices. Machine Learning (or ML) enables AI to self-learn and respond, using data it is exposed to (just like a human brain). Risk management is about using wide ranging perspectives for providing insight, oversight and challenge to bring balance to the efforts of the bank in driving towards its strategic goals. Banks and Financial Institutions are driving transformation initiatives to strengthen Risk Management performance whilst achieving efficiency; and AI may well prove a game-changer here.
In a financial institution, everyone has a responsibility for the management of risk in their day-to-day activities. Business functions (typically designated the first line of defence – FLoD) own and manage their risks through processes and controls; with insights, oversight and challenge from Risk Management (typically designated the second line of defence – SLoD). AI could be deployed across the lines of defence covering various risk management activities.
Just as AI will evolve from mundane and repetitive tasks to more complex pattern recognition for decision making, the evolution of AI within Risk Management is likely to span across three phases. Initially, AI may be used to gather valuable insights for deeper understanding of Risk to enable better decisions. This could be followed by usage of AI in Risk Oversight for providing frameworks, standards and controls within which risk activities can be executed, and to inspect specific aspects of risk taking to ensure they comply with specified frameworks and standards, and highlight exceptions for scrutiny including those in AI installations in the Business. Once AI acquires stability and maturity, it could be used to provide an independent point of view on key risk decisions and challenge the FLoD independently. In short, AI could provide insights to aid Risk Managers, gradually evolving to low-end decision making and later to performing independent checks of processes including of AI deployed in the Business. For instance, AI may be utilized to process patterns in data to identify new risks, and alert Risk Managers in diverse areas such as transaction screening for financial crime, client due diligence or shaping credit appetite in real time based on information processed from commodity prices and currency fluctuations.
Of course, Risk managers too will need to evolve. Risk managers will be required mainly for high-end application of judgement such as assessing the sources of data fed to AI; review the evolution of the AI algorithms as well as the results; and for devising and maintaining the frameworks for the management of risk. Whilst this reality is still somewhat distant, AI will supplant some low end roles sooner.
The advantages of applying AI in the Risk domain are many. Speed of delivery, integrating multiple systems to gather and analyse data, and providing consistent insights with reduced human bias to Risk Managers through real time dashboards are obvious ones, albeit aided by the need to streamline data plumbing just to be able to implement AI. The BFSI sector has already started moving in this direction, but it will be years before AI would be in a position to comprehensively help manage risks for a Bank/ FI unsupervised. Second, it would take some time for AI to achieve the expertise of human experience. An experienced Risk professional would flexibly factor in diverse elements like politics, economics and corporate dynamics to make decisions, while it may be challenging to bring such flexibility within AI’s purview. Therefore, patience would be vital to achieve that level of sophistication. The pace of adoption of technology suggests that this evolution will likely be exponential rather than linear beyond an initial inflexion point.
It would be somewhat naïve to assume that AI would eliminate risks for the BFSI sector; rather it would likely transform risk. For one, the success of AI would depend on the relevance and quality of the data sources fed into it. Bad data is as much a risk for AI as a defectively evolving algorithm. In addition, technology, no matter how robust, does have an element of vulnerability to cyber threats. Films have shown myriad doomsday scenarios of machines going rogue (take your pick!). But immediate concerns centre around the unintended consequences of computer programs reaching bizarre conclusions through defectively re-writing their own algorithms, exposing the institution and its clients to unplanned risk.
If we get it right, AI can transform Risk Management with incisive analytics and cognitive computing beyond the realm of human capacity. If AI is founded on defective data or programs, it could result in perverse results, potentially triggering serious consequences yet to be seen.
What does the future hold? Maybe only AI can tell…
Authored by Rajesh Jogi, Head – Risk Hub India & APAC Risk at Royal Bank of Scotland