Will AI Replace Your Risk Management Analytics?

For the past few decades, analytics for risk management has meant business intelligence (BI) and predictive analytics, which for the current purpose I will also take to include optimisation, a.k.a. prescriptive analytics. Now the buzz around artificial intelligence, enhanced by popular myths around robots (thanks, Hollywood!), has created some confusion. Does AI replaces previous technologies, or work with them? How will AI help us in risk management?

There is no denying that AI and machine-learning technologies have piqued public interest. Impressive developments such as Google’s AlphaGo computer beating the Go world champion have captured everyone’s attention.

In its original and basic form, AI is best described as a computerised system that has the ability to react to changes in its surroundings in an intelligent way, where “intelligence" can be measured against some form of objective function. The applications of such systems can either be broad or focused on a single task, albeit most likely a complex one. A straightforward example is playing a game like Go or chess, because this requires the machine to be able to recognise patterns and assess different competing alternatives.

Nevertheless, whilst there is certainly great potential in AI, it will not sweep away the other technologies used in risk management.

BI Is Smarter Than Ever

Let’s start by looking at BI, a set of tools that makes sense of raw data to help business managers recognise patterns and make decisions based on past performance and trends. Traditional BI tools extract data collected from previous events, such as customer transactions, and produce synthesised information that executives can more easily and usefully work with.

However, BI has evolved significantly since its introduction. Improvements in computing power, data storage, and availability have opened up a whole new range of opportunities for BI, including real-time data analysis. A great example is data streaming, from jet engines or cars – the Internet of Things provides data that enables real-time adaptation, which effectively minimises risks, as you can respond to any issue or failure in real-time and avoid unnecessary escalation.

An area where BI performs exceptionally well is in conduct and compliance risk. BI allows businesses to understand and monitor difficult processes, like payment protection insurance and originations. For instance, many financial services organisations benefit greatly from using BI tools to monitor customer and agent behaviour during collections procedures. Speech recognition technology, one of BI’s more advanced capabilities, enables managers to monitor customer calls in real time, checking interactions between the agent and the customer for compliance and performance purposes. Having access to this data helps businesses ensure that agents remain compliant and follow the required procedures, such as reading a customer their rights at the beginning of a call. This is smart, but it’s not AI, just a good application of BI.

Predictive Analytics Meets AI

Risk managers have used predictive analytics for decades now, going back to the origins of credit scoring in the 1950s. While credit risk modelling remains a foundational technology, the technology is found throughout financial services, not just at to assess customer behavior but to derive more powerful strategies. Optimisation solutions help businesses drive efficiencies and profits, improve customer satisfaction, and manage risk.

Where BI tells you what has happened and what is happening, predictive analytics can tell you what is likely to happen in the future based on the data available. For example, when a valuable customer calls to close her bank account, optimisation analytics can create a sequence of responses to help give the service agent the best chance of retaining that customer’s business. In the area of fraud detection, neural networks, link analysis and other technologies perform complex calculations in real time to determine whether a transaction is out of pattern – for the customer, for their segment, for the merchant or medical provider, for the device, etc.

The best results are achieved when BI and predictive analytics are used in conjunction. For example, an organisation hoping to improve its pricing strategy could leverage BI to view past and current pricing trends and inform the new proposition, then use predictive analytics to identify consumer sensitivity to such price changes, as well as the likely uptake on the new offers. 

The same is true for AI. In a risk management context, AI is often used as an alternative term for machine learning, which is just a subset of the broader topic. Machine-driven tools provide the most benefits when used with other analytic technologies. AI tends to be a black-box technology, making it hard to deliver reason codes and to demonstrate to regulators that customers are being treated fairly by a model. But AI tools can be used to identify characteristics or to help with generating complex features that go into a more conventional predictive analytics model. This hybrid approach enables us to use the power of AI without sacrificing the transparency and palatability of our existing predictive analytics models.

Say we are looking at a banking customer's application, using a risk assessment model. An AI-based feature can be included to review elements such as unstructured text. Using this complex feature and taking into account complimentary data about that customer's past behaviour, the model can do a better job assessing the applicant’s risk.  If the model relied solely on predictive analytics, the information held in the unstructured text could have been missed.

Innovative risk management tools will continue to improve as analytics solutions develop, and it is essential that business managers understand their importance and trust their output. However, AI does not replace BI or predictive analytics tools, rather it enhances them. By using these technologies together, businesses can build more robust models that can help managers to make better informed decisions and minimise risk.

Dr. Andrew Jennings, Senior VP of Scores and Analytics, FICO.

• Read more under:
• BankingTech
• Digital Banking