FinTechTalk: Revolutionising credit and portfolio management  

On 22 October 2024, FinTechTalk host Charles Orton-Jones was joined by Tom Byrne, Head of Product and Innovation EMEA, nCino; Brad Morris,Freelance credit consultant; and Quresh Kolhapurwala, Head of Credit Risk, First Abu Dhabi Bank (FAB). 

Views on news 

The disconnect between financial institutions and the changing conditions of their markets has left many of them with blind spots – aspects of their operations for which they have insufficient or insufficiently timely data. These blind spots can be areas that might benefit from closer risk management, as well as overlooked opportunities to seize the competitive or operational advantage. There are five critical areas where more effective use of data can make a difference: interest rate risk, liquidity risk, credit risk, sales diagnostics and merger integration. Legacy banking has been rather manual from the underwriting to the monitoring the back book.

While automated decision making in the western world has become the norm, the question is now how subsequent processes can be automated as well, such as cashflow balances and customer behaviour. Ideally, banks would like to have predictive analytics bolt onto credit scoring systems to get better terms, limits and better monetary flexibility.

Meanwhile, on the ground, banks are still struggling with data administration, cutting and pasting from annual reports and disparate systems. Availably of data is usually not the problem for big banks with outdated procedures but structuring data and presenting insights to human decision makers are. In many banks, data still sits in excel sheets where this valuable resource remains untapped.  

Moving from reactive to proactive risk management 

In a heavily globalised world, risks are intertwined as well. The traditional way of adjusting to this has been changing your model, which, thanks to digital technology is now being replaced with scenario analysis and the forecasting of capabilities that can help absorb uncertainties such as geopolitical tensions and macroeconomic changes. Adjusting customer credit limits, however, can interfere with the customer experience.

Another problem is the shortage of talent who are experts on both credit and statistical analysis – two huge areas of expertise requiring very different skillsets. As regulatory requirements don’t allow the full automation of credit processes even where data is available, the biggest challenge at this stage is how the data that the AI tool arrived at can be presented to human decision makers and how thresholds can be set above which looming risks must be dealt with. In this scenario, AI is used to take out the complexity of relying on many data sources. Automation, then, can be pushed forward at a later stage as regulator appetite and the technology mature.  

Another area where AI can help is optimising your portfolio, which doesn’t mean moving credit limits for individual customers but, rather, optimising metrics for whole industries. To stress test customers that present higher risk, prior to a lease or a property loan they can be lent money on a 6 month time horizon while the lender can pull in qualitative and quantitative data to see whether they these individuals should be eligible for riskier longtime loans and leases. The fact that AI is gaining ground in credit scoring is demonstrated by the fact that we are witnessing a reduction in the number of underwriters and credit analysts – although it mostly affects junior positions and leads to a higher level of specialisation for senior staff.

As a result, a deep knowledge of a geographical area, industry or a sector can increasingly add a lot of value to a credit risk analyst’s skillset. For institutions to see if they do a good job at managing credit risk, they should look at their days sales outstanding and risk grading to assess the health of their portfolio. Models must be tuned and calibrated to understand the market (barriers to entry, credit requirements) and the qualitative aspects of the business, as well as how they strengthen or dilute the credit risk a business is taking. The number of defaults (NPLs) and arrears are other metrics that can reveal whether risk is managed properly.  

The panellists’ insights 

• A lot of the game today is about understanding your customers better and finding ways of proactively supporting them.  
• Credit experts who want to improve their data skills can familiarise themselves with the Altman-Z score that is not only about credit scoring but also shows the most important parts between a set of accounts that can predict the financial health of an individual borrower.  
• The major measure of success in the long term is whether the operating model is moving from a reactive risk management process where humans do a lot of admin work and reviews take place in annual cycles to data-driven decisions where analytics alert portfolio managers to risks and opportunities.