The revolution arrives: How gen AI is poised to transform banking

There’s an old saying: “You wait ages for a bus, and then two come along at once.” Advancements in technology can be like that – except with exponential growth. Instead of two buses, a whole superhighway of them arrives at once. This illustrates the current state of artificial intelligence (AI).

AI isn’t new. But the past year has brought an explosion of AI innovation – notably, the emergence of generative AI (gen AI). AI is suddenly not only widely available to non-technical users; it’s a hot topic in bar rooms and board rooms as individuals and organizations look to capitalize on its transformative potential.

Nowhere is this truer than in the banking sector. An Accenture report projects that banks will benefit from the highest gen AI productivity gains of any US industry. McKinsey estimates that banking globally could realize value of between $200 billion and $340 billion from gen AI. And a survey by ACFE and SAS found that banking is ahead of other industries in both current AI use to combat fraud and anticipated adoption in the near future.

In recent months, the pace of change has shifted perceptibly from brisk to warp speed. Those who are slow to get on this bus risk becoming roadkill. Or at least competitively disadvantaged.

The gen AI trifecta

There are three core gen AI technologies: large language models (LLMs), synthetic data generation and digital twins. LLMs and synthetic data generation are already well-established in banking; the use of digital twins is more nuanced. Let’s take a closer look at each of these technologies and how they are being applied in financial services.

Living large

A large language model is a machine learning model that can generate text, converse with users, and process and identify complex relationships in natural language.

LLMs have many uses within financial services. Banks use them to power their digital customer assistants, improve lending decisions by boosting real-time credit risk analysis, and thwart fraudsters through improved detection of anomalous behavior patterns.

But banks aren’t the only ones embracing gen AI. Criminals are using LLMs, as well, to fabricate  deep fakes to trick automated identification and verification systems and perpetuate phishing and other scams. For banks, gen AI isn’t just a competitive advantage; in the technology arms race against bad actors, it’s becoming table stakes for fighting fraud and financial crimes.

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. This is a major leap forward for banking, for two reasons.

First, there are critical areas of financial services where data could be improved or supplemented – in the realms of climate and transition risk, for example, where synthetic data can help banks and insurers model environmental uncertainties, where historical data falls short of predicting erratic and more catastrophic natural disasters and climate patterns.

Second is the need to safely handle personally identifiable information (PII) and comply with regulatory requirements. Banks hate being in the news for the wrong reasons. None want to be fined for breaching prescribed privacy and security standards or, worse, violate their customers’ trust.

Synthetic data generation allows firms to create artificial data sets that replicate the nuance and diversity of real-world data. This gives them latitude to train and test their models with less risk and can also help them reduce potential data bias.

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 using fewer physical assets in day-to-day operations, some believe digital twins are of limited use.

Don’t dismiss digital twins entirely, though. It’s true that the Internet of Things (IoT) has always been more relevant in industries such as manufacturing, energy and utilities, and retail than in banking. But IoT can be in the conversation for banking, too.

With IoT data intrinsic to creating a digital twin, one can make a case for using twinning in banking. ATMs aren’t going away any time soon, and they are one example where a digital twin approach could prove valuable.

Putting gen AI to work

From front office to back office, every part of the bank will be affected as generative AI use cases are tested and deployed. Its potential uses extend across the business, including:

• Fighting fraud and financial crime. As noted previously, gen AI will help banks improve real-time fraud detection through transaction pattern analysis (across millions of transactions per second) and suspicious behavior flagging. Firms can also use synthetic data to conduct penetration testing on existing fraud control systems, then fine-tune to optimize defensive capabilities. On the anti-money laundering (AML) front, LLMs can help AML professionals fine-tune their transaction monitoring alerts and reduce false positives.

• Enhanced credit decisioning. Gen AI will enable banks streamline their loan processes and support faster cash-in-hand to customers. The effective use of synthetic data, for example, can help banks find and eliminate decisioning bias and improve the overall accuracy of their credit risk scoring models, while LLMs can automate credit scoring.
• Real-time risk assessment and monitoring. Gen  AI will help banks develop better risk assessment models and support better scenario modeling to evaluate potential risks. Synthetic data generation, for example, can enable banks to simulate once-in-a-generation Black Swan events using sparse data sets.
• Increasing operational efficiency. Gen AI-powered automation will streamline processes, reduce errors and lower operational costs. In particular, LLMs can help banks automate and expedite manual and data-heavy processes like customer onboarding, loan processing and fraud investigation, just to name a few.
• Improved customer experience. Gen AI-powered chatbots, apps and money services will provide customers with more human-like digital interaction and services. LLMs have the potential to help banks improve their complaints management processes across channels by decreasing the average response times while increasing the volume of complaints handled.
• Personalized customer interactions. Gen AI will help banks analyze customer data to personalize product offerings and recommendations. It may also eventually be used to provide investment advice and wealth management services.
• Market trend analysis and prediction. Gen AI will enable banks to analyze and predict market trends and provide insights to support decision-making and improved strategic planning. It can also enhance economic and market condition simulations.

Ethical AI: A necessary emphasis

Banking in 2035: Three Possible Futures, a report from Economist Impact and SAS, noted that “AI has the potential to perpetuate discrimination and exclusion due to the biases held by its developers.” As such, trustworthy AI is essential. Banks must employ AI ethically.

As banks infuse gen AI into their operational models, they must maintain a focus on consumer protection, data privacy and security, and eliminating bias from AI decision-making. While AI can deliver extraordinary power to banks and financial institutions, that power comes with significant responsibility.

A modern-day revolution

None of us were around for the Industrial Revolution. But it’s not hyperbole to say that we’re likely witnessing the beginnings of a similarly significant technological revolution right now.

Yes, there are questions to be answered, issues to work through. But, placed in the right hands and deployed in the right ways, gen AI has tremendous potential to reshape banking’s “business as usual” and forge a more inclusive and equitable financial ecosystem for all.

Alex Kwiatkowski is Director of Global Financial Services and Julie Muckleroy is Global Banking Strategist at SAS.