By Lee Thorpe, Head of Risk Business Solutions, SAS UK & Ireland
With Brexit just around the corner, the European banking sector is under significant pressure to forecast and mitigate potential economic shocks. Given the levels of uncertainty over possible outcomes whilst the details of any potential deal (or no-deal) still to be confirmed, how forewarned can banks actually be without testing hundreds of possible permutations?
Following on from the latest European Banking Authority stress test, in which continental banks generally fared better than their British counterparts, financial institutions must continue to ensure that they understand the consequences of a wide range of possible scenarios. This will facilitate operational planning and may provide mitigation strategies for the more probable or consistent impacts of Brexit. The process needs to be automated to constantly update crisis responses using analytically updating scenarios so they can react to any market restrictions that result in an economic downturn just as quickly as customers and the markets react.
The business benefit of analytics
Having to organise and invest for regulatory measures may not be a preferred business priority for financial institutions, but there is a significant upside to getting a better understanding of the market influence on net interest margin and capital, especially during a crisis. The automation of processes to execute complex analytical models means that many different scenarios can rapidly be put to the test to help support strategic decisions or ensure preventive measures are implemented swiftly.
We’ve come a long way since the 2008 crash. Artificial Intelligence (AI) is starting to be utilised in the banking world although less so within Risk and regulated processes. AI can help risk teams map highly complex scenarios and approximate potential outcomes focussing fully production processes or guiding where humans should focus analysis on to make an informed data driven decision. Banks need to collaborate with AI to build the strongest possible understanding, informed by deep analytical insight.
The biggest test is still to come in March – banks must go beyond the regulatory requirements and ensure they have the foresight to prepare for multiple plausible outcomes, even while hoping for a deal to be struck.
Making analytics a reality
With all of that said, how can banks actually implement an effective analytics function to form the basis for resilient risk management? What are the cultural changes that must be put in place, and how should risk leaders go about communicating the value of advanced analytics to the wider organisation?
Here are three key areas to consider:
1. Set a vision with clear expectations of longer-term benefits
It goes without saying that senior managers need to be behind any analytics investment, but it is not enough for them simply to decide that they want the organisation to use advanced analytics or declare that it is now data-driven. An analytics strategy must be set within a clear vision for the company’s future, business direction and goals.
Many banks fail to generate value from their analytics investment because they focus on the wrong problems or projects. It is essential that any major investments in analytics are focused where they can add maximum value, with the recognition that this may not be where project budget is available to fight the latest fire. Potential use cases should be carefully assessed to make sure that they are focused on strategic issues that are data intensive and complex or require fast reactions.
2. Get the basics right
Without the correct foundations, analytical resource will be frustrated. Initially, using the buzz words to create an expectation and facilitate budget may not be what you need. In general, for large companies, automating processes and improving access to existing tools and data can provide a significant uplift in efficiency.
Many banks have evolved independent processes with manual interactions. Supposed analysts spend a large proportion of time ensuring that these processes run effectively, so fixing them and releasing resource to add value has tremendous benefits.
Does the problem require a complex automated solution? Many regulatory processes that define capital requirements or are associated with customer outcomes often aren’t suitable for AI. Capital can’t change without rigorous model approvals – and a self-improving algorithm is unlikely to be approved by the regulator. Also, a key customer interaction for a mortgage, say, needs to be fully understood. If the AI approved them one day and declined them the next, it wouldn’t be fair treatment if the process couldn’t be fully explained in a transparent manner.
Finally, are you increasing model risk? The senior managers regime is there to ensure executive responsibilities cannot be delegated. Self-learning processes may change over time. Are bank executives fully capable of understanding all the risks of these complex process now and as they change over time?
AI should be used in the correct use case with appropriate feedback mechanisms for self-improvement. Alternatively, an auto-updating process can be created based on unclean or biased data.
3. Change culture with new tools and data
If you decide advanced analytics is needed and the use case is appropriate, simply providing new tools will not improve results. Clean data is essential for the process to deliver value.
With new tools and data the way problems have traditionally been solved can be replicated on a faster infrastructure with performance improvements. But for real value to be delivered cultural change is required to ensure that teams think and interact differently. This is the hardest to deliver and requires senior buy-in to push through change that is only enabled by new tools and data sources.
Simply changing someone’s job title to ‘data scientist’ doesn’t improve productivity but is likely to increase salary expectations. Failure should be expected, without giving employees the space to experiment.
The art of the possible should be understood by both business and IT with business objectives aligned with and informed by IT strategy. It’s key for IT to deliver an integrated infrastructure that allows processes to change and advanced models to be quickly deployed to execute and achieve better risk management or customer outcomes.
Conclusion
There’s plenty of scope for the banking sector to prepare itself adequately for the next raft of challenges. With the advent of AI, risk analytics capabilities are becoming more advanced. Imagine having a whole department of expert risk analysts working at lightning speed – that’s what a well-implemented intelligent analytics deployment could look like for your company. With the international outlook currently uncertain, both economically and politically (and when isn’t it?), banks have a responsibility to ensure they have the clearest possible view of the many potential futures they face.
The final thing to note is that AI, long held up as a magic bullet, is finally making its exit from the high-tech companies and moving into the traditional boardroom. It can be done, and it can have a huge impact on the accuracy and detail available to risk analysts. But good data and advanced analytics has to be at its heart. AI without good data, tools and culture is nothing. Banks can effectively analyse the risks they face – but they must act now.