Banks must refresh AI and Machine Learning models to reflect recent economic turmoil

How enhancing your modelling capabilities can help your bank navigate economic instability and the cost-of-living crisis

The past few months have been a rollercoaster for the UK economy. Even before the recent upheavals at Number 10 and the Treasury, experts were forecasting a 40-year high for inflation—a rise driven by the aftermath of the pandemic, Brexit, the global energy crisis, and the ongoing war in Ukraine. Meanwhile, the Bank of England has just raised interest rates by the biggest jump since 1989 and predicted the longest recession in 100 years, while house prices have fallen three times in the past four months. 

As the economy contracts over the coming months, it’s going to hit both households and businesses where it hurts. Higher rates will mean higher mortgage repayments, cutting into discretionary spending and pushing some borrowers—through no fault of their own—into vulnerable territory. As consumers tighten their belts and overheads increase, the outlook for businesses is grim too, particularly for smaller shops, restaurants, and entertainment venues.

To support customers during these challenging times and help the UK economy find a quick path out of recession, banks have a key role to play. However, their ability to help will depend on their ability to model, analyse, and predict the financial situation, outlook and needs of each household and each business they serve. 

At the same time, banks need to protect themselves from bad debt—according to recent statements from Lloyds and NatWest, their risk models have already indicated that they will need to set hundreds of millions of pounds aside to cover loans that they predict won’t be repaid.

The good news is that most banks are in a relatively strong position to weather this situation. Capital reserves are high because funds set aside during COVID are largely still intact. Meanwhile, higher interest rates will make it possible to make money on net interest margin for the first time in years. So, banks have both the resources and the motivation to play an active role in the UK’s recovery.

Still, banks need to be careful. If the past couple of years have taught us anything, it’s that the world is far more unpredictable than we thought. Post-pandemic, we’ve seen that many of the models that banks use to create forecasts, make predictions and support decision-making are no longer accurate—and that’s a big concern. If you don’t understand credit risk at an individual customer level, how can you lend responsibly? If you don’t know who your vulnerable customers are, how can you help them stay financially secure? If you can’t target your collections processes effectively, how can you avoid a tidal wave of bad debt?

The failure of these models is particularly troubling because during the pandemic, most banks put their model development and infrastructure projects on the back burner. Many haven’t updated their modelling platforms for a decade or more. At the same time, there has been a huge increase in reliance upon AI and machine learning models to inform banks’ decision-making across the board.

It’s clear that the regulators have recognised the problem this presents with regard to controlling the risks inherent in models. We’ve seen a deluge of discussion and consultation papers from the PRA, such as DP5/22 on the use of artificial intelligence and machine learning, and CP6/22 on model risk management. And initiatives such as the FCA’s Consumer Duty policy will clearly require more accurate, detailed and intelligent analytics to ensure banks treat their customers fairly and provide products that deliver genuine value.

To navigate an unpredictable future, banks will need new and better models across all areas of the business—from identifying vulnerable customers and creating collections strategies that minimise bad debt, to managing credit risk and helping UK businesses find a path out of recession. 

At the same time, enhanced modelling platforms and capabilities can help banks take advantage of opportunities and put their money to work, while reducing the amount of capital they need to keep in reserve to offset unexpected risks.

Banks need to accelerate the model lifecycle, so they can rebuild and recalibrate models (whatever the coding language) faster to meet today’s challenges—whilst enabling compliance with current and emerging regulations by virtue of a governed and explainable framework.