From evolution to revolution: The transformative effect of generative AI in banking

As the old saying goes, "You wait ages for a bus and then two [or possibly three] come along at once."  This saying can be updated to reflect life in our increasingly digital world: "You wait ages for a genuine disruptive technology and then two [or possibly three]arrive simultaneously."

This phrase succinctly captures the transformative effects of generative AI (GenAI) and its key underlying elements: Large language models (LLMs), synthetic data generation, and digital twins.

A report from Accenture highlights that banks are more likely to benefit from the highest GenAI productivity gains than any other industry in the US. McKinsey estimates that banking has an annual potential increase in GenAI-driven value of between $200 billion and $340 billion (nine to 15 percent of operating profits) as a result of improved productivity (with the most significant gains coming in the sub-sectors of retail and corporate), it’s clear that GenAI is a genuine game-changer like no other.

HFS Research, a respected independent analyst firm, surveyed enterprise leaders and found that by 2025, GenAI’s impact will be bigger than every other major technological breakthrough in history, surpassing the printing press, steam engine, the internet, and the smartphone. For something that’s only existed for a relatively short time, that’s quite an accolade. And all available evidence points to this being reality rather than hyperbole.

In a few months, the pace of change has shifted perceptibly from brisk to warp speed. Failure to either buckle up or get on board at the beginning has put those who are hesitant, scared, uninformed, or downright dismissive of GenAI at a distinct disadvantage.

Deployment has already begun within forward-thinking financial firms, and the benefits are already being reaped. Not only are significant developments occurring faster, but they are also having a significant positive influence on operational capabilities. In the blink of an eye, we’ve transitioned from an evolutionary period – which SAS’ Julie Muckleroy helpfully outlined in a recent blog – to a revolutionary epoch. Viva la GenAI revolución!

History has shown us that a successful overthrow requires speed, strength, determination, planning, capability, resources, an arsenal of weaponry, and a ruthless willingness to use said weapons to accomplish the objective/s. While GenAI’s revolution is thankfully bloodless, the ordnance involved is incredibly powerful.

Placed in the right hands and deployed in the right way, GenAI can radically reshape banking’s “business as usual” activities in lending, fraud and financial crime detection and prevention, and customer transactions and interactions. Of the three weapons mentioned earlier – LLMs, synthetic data generation, and digital twins – each has applicability and merits.

The suitability of LLMs and synthetic data generation in banking is clear, whereas the use of digital twins is more nuanced. Subsequent blogs will deconstruct each element in detail, but let’s begin with an overview of each.

Do you want to go large?
An LLM is a machine learning model that can process and identify complex relationships in natural language, generate text, and converse with users. LLMs are proliferating rapidly, driven by developments in the open-source community and big tech corporations. They go from nowhere to everywhere in the same amount of time it takes the Earth to travel around the sun.

LLMs have broad applicability within banking, from the obvious – such as a digital customer assistant delivering a hyper-personalized service – to the subtle, for example, elevating banks’ credit analysis and fraud prevention capabilities. LLMs can improve the effectiveness of lending by bolstering real-time risk assessment capabilities and thwarting fraudsters through improved anomalous behavior pattern detection.

However, where the latter is concerned, it’s important to acknowledge that it’s not just banks leaning in on GenAI. Financial criminals are using technological advancements in AI for nefarious purposes, such as creating images or voice cloning that can trick automated identification and verification systems into thinking that the purported customer is legitimate when they’re not. Consequently, banks need to be more vigilant than ever when preventing financial losses through fraud.

Solving challenges through the science of synthetics
Synthetic data generation refers to on-demand, self-service, or automated data generated by algorithms or rules rather than collected from the real world. For the banking industry, this is a highly valuable – some might say vital – leap forward for two fundamental reasons.

First, there are mission-critical areas of banking where data could be improved or improved, but the consequences are severe. Examples include but are not limited to accurately assessing climate-related risks (a topic that’s rising ever higher in importance), identifying payment fraud (where millions of transactions are taking place every second), and lending prudently to customers of different types and sizes (given lending is banking’s raison d’etre).

Second, the issue of safely handling sensitive personally identifiable information (PII) within the permitted regulatory compliance parameters. Banks hate being in the news for the wrong reasons, and there have been too many instances where fines have been levied for breaching prescribed standards. Becoming better is a worthy objective for every entity or individual, and the banking industry certainly is not immune to the need for continuous improvement.

How can synthetic data help? In the sphere of fraud and financial crime detection and prevention, banks can use this GenAI element to extrapolate from rare events and anomalies to train robust models on specific fraud and anti-money laundering topologies. Institutions can also use synthetic data to conduct penetration testing on existing fraud control systems before fine-tuning as required for optimized defensive capabilities.

Moving from fraud to risk, the use cases for synthetic data generation are equally notable. There’s the ability to simulate once-in-a-generation Black Swan events using sparse data sets, the ability to train models on new exogenous developments such as climate change, an opportunity to enhance micro/macro-economic and market condition simulations, and to improve the accuracy of the complex models used for credit risk scoring.

With customer intelligence serving as the third part of the trinity alongside fraud and risk management, synthetic data plays a key role, too. Banks can create personalized services without using sensitive PII and instead use synthetic behavioral profiles to develop offerings confidently and safely, knowing that no breaches will occur. Finally, synthetic data can be used as a critical tenet in training models for customer acquisition and ongoing marketing.

Digital twins and banking: Separated at birth?
Digital twins are virtual models of real-life objects or systems built from historical, real-world, synthetic data or a system’s feedback loop. With banks reducing – but not eradicating – the number of physical assets used in day-to-day operations, some believe digital twins are of limited use. But is there a counterpoint to this argument?

Before we dismiss digital twins entirely, let’s understand whether there’s a case to be made for their recurring use in pursuit of performance improvements. The Internet of Things (IoT) has always been of peripheral relevance to banking, with industries such as manufacturing, energy and utilities, and retail making far greater use of connected devices for continuous monitoring.

However, that’s not to say banking is missing from an IoT conversation. With IoT data intrinsic to creating a digital twin, there is a case to be made for using twinning in banking. ATMs remain prevalent, will continue to exist, and represent a relevant example of where a digital twin approach could be taken.

Take it away, Paul and John
“You say you want a revolution,
Well, you know,
We all want to change the world…”

… sang The Beatles. In GenAI’s case, it’s changing the world of banking in ways that were previously unimaginable and at astonishing speed. This is just the beginning of what undoubtedly will be a fascinating journey. Let’s go!

Alex Kwiatkowski is Director Global Financial Services at SAS.