No, human-like chatbots have not yet replaced loan officers. And today’s investors wouldn’t rely solely on the wisdom of financial robo-advisors. But generative AI (GenAI) has already made its mark on the financial services and banking industries—and it shows no sign of stopping.

In the McKinsey Global Institute’s analysis of GenAI’s potential value across industries, banking ranks as one of the top industries most likely to benefit, with GenAI potentially adding $200 billion to $340 billion in value.

Most of these gains will be achieved through improved employee productivity, especially with GenAI’s ability to create content such as text, images, and more based on inputs or prompts. Consider, for example, Marsh McLennan, a professional services firm with businesses in risk management, investment advisory, and wealth management. The organization recently deployed a proprietary GenAI assistant called LenAI to help employees write and improve their communications, quickly find answers to specific questions, and summarize documents.

Within 30 days of deployment, the tool was used by 15,000 employees, 94% of whom said it improved their productivity and efficiency. Employees reported saving an average of eight hours per week and spending 20% less time on repetitive tasks when using the tool.

Ultimately, GenAI could redefine how banks interact with customers, inspire new risk management strategies, and create new avenues for revenue. For now, though, banks are understandably focused on “incremental innovation,” says Miriam Fernandez, director of AI research and analytical innovation for S&P Global. “It’s about little steps and little gains that together make a big difference, and only with time, they will be transforming their business model more holistically.”

According to a study by UK global management consultancy Oliver Wyman, more than 70% of the GenAI use cases at the financial services organizations surveyed are still in the proof-of-concept or pilot phases.

Moving past the pilots and achieving GenAI advantages will require these organizations to overcome large challenges, including a stringent regulatory environment and skills shortages, while at the same time establishing best practices that promote collaboration and drive adoption.

How GenAI is making a difference in banking today

For decades, banks have been using AI and its sophisticated algorithms to identify patterns in consumer behavior, predict market trends, and drive decision-making. So when OpenAI’s ChatGPT burst onto the scene, quickly followed by GenAI technologies from Google, Meta, Microsoft, and others, “banks were already well positioned to move forward with GenAI,” says Fernandez. Over the past few years, she points out, they’d been investing in machine learning and deep learning to automate their processes and to make predictions based on patterns in their data.

For banks, the timing of GenAI’s entry into the mainstream is fortuitous; many are facing increasing competition from fintechs and a challenging regulatory and compliance environment. GenAI promises to help banks meet these challenges by making staffers more efficient and speeding up time-consuming tasks, such as writing contracts and analyzing investments, while improving the customer experience.

Some of today’s top financial institutions are exploring a growing number of innovative use cases in the following areas:

• Speeding up time-consuming regulatory compliance tasks. In late 2023, the Federal Deposit Insurance Corporation in the United States proposed a set of rules requiring big banks to hold more capital to better withstand risk. Rather than manually examine these complex rules, Citi turned to GenAI to sift through the 1,089-page proposal word by word and arrange the document into digestible chunks and key takeaways. As a result, the company could more quickly assess and present the plan’s effects to its treasury organization.
  
  In today’s ever-evolving regulatory environment, GenAI could play a major role in streamlining the compliance process by monitoring transactions and ensuring that they meet all legal requirements, thereby reducing the risk of costly regulatory violations for banks.
• Synthesizing information to provide faster financial guidance: GenAI could also provide easy access to information that would normally take financial advisors hours to retrieve. Case in point: Morgan Stanley Wealth Management is using technology from OpenAI to synthesize vast amounts of data on everything from capital markets to regions around the world. With the assistance of a chatbot, financial advisors can ask questions about its data and receive immediate responses that can help them better serve their clients.
  
  By leveraging chatbots powered by GenAI, Fernandez says, “Banks can help their employees access information more easily, such as client data or market information.”
• Providing more personalized financial advice: Similarly, JPMorgan Chase recently launched a GenAI-powered tool to support thematic advising, an investment approach that relies on research to identify investments based on developing macroeconomic, geopolitical, or technological trends rather than traditional industry sectors or company profiles. Dubbed IndexGPT, the tool works by generating a list of keywords associated with a new theme, which are then fed into a separate natural language processing model that sifts through news articles to identify companies involved in the space. The resulting tool helps advisors pick, analyze, and recommend investments that are tailored to their customers’ needs.
• Combating fraud: GenAI can serve as a powerful defense against fraud—a crucial functionality given that merchants are expected to lose more than $362 billion globally to payments fraud between 2023 and 2028, according to Juniper Research.
  
  Mastercard, for example, designed its own proprietary GenAI model called Decision Intelligence Pro to detect fraudulent transactions in real time. Trained on nearly one trillion data points, the solution can assess whether a transaction is legitimate in less than 50 milliseconds by understanding the relationships between multiple merchants surrounding a transaction. For example, the GenAI model may determine that a transaction is fraudulent if a cardholder’s history indicates that they’re unlikely to purchase from the business involved.

GenAI challenges for banks to overcome

Advancing along GenAI’s maturity curve requires financial institutions to overcome large organizational, cultural, and technological challenges.

Chief among these is addressing questions about data privacy, security, and the ethical use of AI. “The challenge is particularly demanding for global banks that operate in multiple places. They need to understand the different AI regulations and comply with them, particularly when it comes to data privacy and security,” says Fernandez.

For example, European Union lawmakers have formally adopted the world’s first comprehensive law on artificial intelligence, the AI Act, which aims to protect consumers from potentially dangerous applications of AI. The Cyberspace Administration of China has also put forward its own regulations to govern the use of GenAI services. Global financial institutions will need the talent and expertise necessary to navigate these varying regulatory frameworks.

Beyond regulations, the output of GenAI models can also be problematic. Due to their inherent predictive nature, GenAI models are prone to hallucinations—a phenomenon that occurs when an AI model provides responses that are nonsensical or outright inaccurate. Just recently, Google’s new AI model Gemini began generating historically inaccurate and offensive images, prompting an apology from Google.

Given these risks, some banks are taking a wait-and-see approach to using GenAI in customer-facing applications.

“Conversational AI powered by GenAI is something we still have to work toward,” says Promiti Dutta, Citi’s head of analytics technology and innovation, during a recent talk. “In an industry where every single customer interaction really matters and everything we do has to build trust with our customers, we can’t afford anything going wrong with any interaction.”

Protecting proprietary or sensitive data is another obstacle banks must overcome to balance the risks and rewards of GenAI. For instance, J. Scott Christianson, associate teaching professor and director of the Center for Entrepreneurship and Innovation at the University of Missouri, says, “ChatGPT can be tricked into leaking information that’s collected—hoovering up huge amounts of information that might have your Social Security number or your banking information.”

In early 2023, a popular GenAI-powered chatbot experienced a data breach that partly exposed the chat histories and some users’ personal details, including e-mail addresses and the last four digits of their credit cards. Although the exploit, caused by a bug in an open-source library, was addressed quickly with minimal damage, it prompted some organizations to restrict employees from using the tool due to fear of data leaks.

To minimize the risk of data leakage, Christianson says, “Many banks are exploring ways to develop their own GenAI systems that they can use internally and have the appropriate controls on.” After all, he adds, “There is hesitancy on the part of banks to have employees use these generic consumer-grade AI tools because of the potential for employees to put proprietary information into them.”

How banks can strategize for success

Fortunately, banks can deploy strategies to address issues of compliance, model output, and data security and to ensure that they gain the most value from GenAI today and in the future. For instance, before embarking on a GenAI journey, financial institutions must determine what they hope to get from this technology by identifying value-add use cases. “AI has to align with a bank’s business strategy,” says Fernandez. “Having just AI tools is not enough.”

Other measures are outlined below.

Establish metrics. Once GenAI projects are underway, a set of clear and concise metrics can help keep such initiatives on track. That’s not to suggest, however, that calculating the ROI in GenAI is easy. “Increases in productivity can lead to cost savings and better cost-to-income ratios for banks,” says Fernandez. “But it can be difficult to measure the financial return of an AI initiative. There is also a considerable up-front investment. It can take years for those gains to materialize, and the benefits are not always linear.”

According to research firm Gartner, by 2025, growth in 90% of enterprise deployments of GenAI will slow, as costs exceed value.

Look beyond cost reductions. Christianson says banks must look beyond cost reductions to truly calculate the value of GenAI. “Rather than looking at reducing the bottom line, banks need to think about how they can use this technology for ideation, strategy, and scenario planning,” he says. “These are the areas where C-level executives should be concentrating because the whole cost reduction focus is a race to the bottom.”

Get employees engaged with the technology. People are a crucial component of any GenAI strategy. But while success hinges on employees fully accepting GenAI tools, the fear of job loss can impede progress. A recent report by Goldman Sachs estimates that as many as 300 million full-time jobs globally could be disrupted by AI.

To allay employee fears and drive adoption, Christianson says, “Companies need to be very cautious about this idea that you can just plop GenAI into a workforce and employees will immediately become more productive. We may have to rejigger the way that we do work.”

One way that banks can change the way teams work is by fostering greater collaboration between business and IT when identifying and deploying GenAI applications. For instance, it’s not enough to simply launch a GenAI model that assesses the creditworthiness of customers. Rather, IT must take the time to demonstrate how this tool can help create more accurate credit profiles, thereby minimizing risks for the bank while improving employee performance and overall productivity. Says Christianson, “IT has to work hand in hand with the business to understand how to make these systems more transparent.”

Uplevel technology skills. Banks must also reconsider the composite of their workforce when it comes to deriving value from GenAI initiatives. “Banks need to reskill internally to develop GenAI infrastructure, from models to AI governance and risk controls to internal culture and education,” says Fernandez. Expertise is needed in data science, machine learning, software development, model risk analysis, policy, and governance issues. Yet decades of working with legacy systems has meant banks must play catch-up to bridge significant skills gaps.

However, there are signs of improvement in terms of shoring up the necessary talent. According to a report from Evident, an intelligence company that benchmarks and tracks AI adoption across the financial services sector, the volume of AI talent at the world’s 50 largest banks grew 10% from May to September 2023. Overall headcount, meanwhile, declined 2.5% in that time period.

To maintain momentum, Christianson says banks should position “a futurist on their board who will be rewarded for thinking about GenAI initiatives that may not lead anywhere but can inform everybody about the potential outcomes and guide the board moving forward.”

More GenAI change ahead

These are early days, and financial institutions are only beginning to grapple with issues such as regulations, skills shortages, data security, and leadership.

But the momentum is growing based on the speed and efficiency gains in key aspects of banking, such as regulatory compliance and financial advisory services. This will serve as the foundation for more sophisticated forays into GenAI that will result in new business models and more personalized service.

Realizing GenAI’s full value will require industry leaders to raise important questions about the risks and rewards of this developing technology and determine the necessary best practices to strike a balance between innovation and responsible GenAI application.