GARP and SAS partnered for a survey in 2018 to understand the use of AI & ML models in risk management. Since then, we've seen an increasing demand in the market to credit risk transformation projects (CRTs). Because of that, the survey was expanded in 2022 and has been extended to CRTs to bring insights to the market related to the priorities, the scope and the key success factors of these transformation projects.
In this blog, I wanted to share five key learnings from the new findings in the survey and interpret their likely implications for the financial services industry:
1. 79% said CRTs are medium to high priority compared to other transformations
CRTs are a high priority for financial services companies. About 79% of participants said credit risk transformation is a medium to high priority compared to other transformations. And 76% of participants also noted that CRTs are either in progress or in their short-term agenda.
I’m not surprised by these results considering the conjuncture the financial services companies are experiencing. There are two main drivers of these results: The first is today’s considerably uncertain economic environment. This started with the pandemic and then the supply chain crisis, which we are still experiencing today. This uncertainty has been exasperated by the war in Eastern Europe and the cost of living crisis experienced in many countries, including the developed ones. These turbulent times have forced business and risk leaders to search for a more resilient credit risk management framework.
The other driver has been the ongoing advances in digitalization in banking over the last five to 10 years, mainly focusing on providing the customer experience the customers are used to receiving in other industries, such as mobile phones or online shopping. The financial services industry has tried to catch up with this trend but mainly focuses on the front end, customer service or marketing. We are observing that, at the back end, credit risk infrastructures are having a higher priority now to catch up with the experience at the front end.
2. Siloed structure slows down the transformations
Although CRTs have been ongoing for the last 4-5 years, the survey results show that only two percent are completed and only three percent of the expected return on investment (ROI) is achieved.
I interpret this as such: because of the siloed structure of financial service companies, CRTs are typically initiated to meet the needs of one part of the organisation in a responsive nature rather than a proactive approach. CRTs will deliver suboptimal results without a big-picture view of the end-to-end customer lifecycle requirements. For example, in retail banking, we’re observing that while organisations may be transforming their data structure, decisioning systems or strategies, collections capabilities are not part of this transformation. This is partially because of different priorities and because they run on different systems. This makes the transformation efforts ongoing, completed in one part of the organization and starting in another. This is relatively inefficient and costly.
3. The most sought-after capability is optimization
About 72% of the participants said their business objectives include optimizing credit decision-making.
I’m not surprised to see optimization having the highest % in business objectives. This aligns with what we at SAS have discussed with many of our clients, especially the major multinational financial institutions. These companies have matured in using predictive analytics in their decisions, from originations to collections. They are now demanding to bring prescriptive analytics benefits into their decisioning ecosystem.
This is helping both the bank’s marketing teams’ personalized customer experience aspirations and the credit risk management teams’ differentiated treatment requirements. Banks that successfully implement optimization can balance risk and reward more effectively, as more than one business objective can be optimized in one optimization model. For example, suppose the bank is using optimization modelling to identify the right loan amount for the customer. The model's objectives can be chosen to minimize default risk and maximize revenue while using a profit model.
4. Better data management is still a concern
About 53% of the participants said their transformation objectives include better data management, another result from the survey that I had a chance to validate with my client conversations. Several multinational banks today are building ‘data transformation’ teams working in parallel to their business-as-usual (BAU) data teams. Transformation teams focus on bringing connected databases to the team, democratizing access to data and enriching the data content. In contrast, BAU teams focus on responding to day-to-day data-related business and IT requirements.
Data is at the core of every decision in banking. The data the customers share with a bank, explicitly or implicitly, is an asset. Customers expect the banks to use this asset to benefit them, providing them with better borrowing experiences and products. Customers expect data-driven offers and experiences similar to those from companies like Amazon, Google or Apple. That’s why financial services companies are trying to develop connected and enriched databases with a 360 view of their customers.
There are two common data enrichment trends: The use of alternative data and the use of open banking. Banks can observe significant benefits in several use cases of alternative data. For example, it can be used to increase the effectiveness of fraud prevention rules, mainly to decrease false positive ratios. It can also be used to identify and acquire new customers, for instance, no file / thin files where credit bureau data is limited. It can also retain more customers by developing more precise behavioral models. Many FinTech companies are already using such models to enhance customer experience. My favourites are buy now, pay later (BNPL) providers like Affirm from the US or Klarna of Sweden, where customer experience is at the core of their business model.
5. One key challenge is the expanded use of AI
When we asked what the most challenging area of transformation is, most participants (48%) said it’s using AI.
AI and ML are modelling approaches SAS has been discussing for many years, and with discussions being accelerated with the pandemic and cost of living crisis. However, especially for banking, it’s still a dilemma to use ML vs. traditional models due to various known reasons. In banks, ML models are mainly used to improve the predictive power of scoring or to process alternative data to determine creditworthiness.
FinTechs have been more successful in mastering ML models mainly in digital lending.
For example, Klarna, a Swedish company, uses ML models to predict customer payment behaviour. Another common use case is to streamline and automate lending processes. For example, Upstart, a hybrid lender from the US, which lends some loans directly and facilities loans for other lenders, is one of the most high-profile start-up companies using AI to streamline the loan process. They can process loan applications end-to-end online via a smooth customer onboarding experience. ZestFinance is another FinTech using ML to process alternative data to collect information on “thin file borrowers.” They had started by targeting student loans, so they specialized in thin files.
Amazon has another successful example of small business lending. The company has a vast amount of proprietary information on what products are sold on its website, how customers feel about those products and the financial status of the companies that make them. They are using this data in an ML model to target companies for small business loans. Last year, they lent out roughly $1 billion to small businesses with this lending model.
The gap between banking and fintech capabilities in terms of the use of AI is narrowing. However, the survey shows it’s still a concern. This is mainly due to the lack of AI-powered technologies.
Financial services will modernize with AI integration
Ultimately, I am aware of innovative products and services projects at financial services companies waiting to enrich available customer data and leverage AI models. The most significant enabler of these capabilities is the underlying technologies, which are the roadblocks for risk transformation projects. 68% of survey participants said modernization of IT infrastructure is one of the transformation objectives. Most current requirements from global companies involving ‘future-proof,’ ‘cloud native,’ agile and integrated technologies are another proof of that. Business leaders are eager to bring new products and services based on advanced capabilities – such as 360 views of customer data or connected decisioning – and as a response, IT leaders are looking for more integrated and flexible systems. The financial services companies must move from product/solution-based technologies to connected, single platform-based credit risk decisioning technologies. By adopting such technologies, it will be possible to implement more considerable scope changes in shorter times.
