3 Promising (but Real) Uses of AI in Retail Banking

by Frank H Bria

According to a McKinsey report titled Digital America: The Haves and The Have Mores, financial services placed fourth on a list of the most digitized industries. One of the key technologies driving this digitization is Artificial Intelligence (AI). A broad area that encompasses sub-disciplines such as Machine Learning, Deep Learning, and others. AI, in many use cases, has demonstrated tremendous effectiveness in delivering cost saving and revenue-boosting capabilities to retail banks. This promise has been slow in unfolding, but we are now beginning to see banks experimenting with and implementing both generative and traditional AI to address efficiency issues and boost productivity and profitability. However, even though AI in finance is a dynamically growing trend, there are hardly any solutions that rely on the technology fully. AI is still treated more as a component of a solution, not a solution itself.

Another challenge that traditional banking and finance faces is that while consumer trends and expectations are a major innovation driver in the retail banking sector, disruption by challenger fintech startups is slowing down rapid adoption of cutting edge technologies.

Digital-only banks like Chime, Monzo, and N26, although in some cases experiencing significant disruptions to their operations, have created digital-only products that appeal to an increasingly digitized customer base. These entities may also be better positioned to integrate AI tools into their digital-native infrastructure, giving them a first-mover advantage as AI technologies mature. Such pressures have compelled legacy banks to create in-house innovation hubs and corporate venture capital arms to find ways to disrupt their businesses. While they certainly have the capital and the resources to do this, the question remains: is acting as a venture capital arm of your own technology division really the most effective way of getting new technology into the hands of your customers? However the answer to that question pans out over the next few years, one thing is clear: as a core technology in this wave of disruption, AI is squarely on the radar of legacy banks as a prime candidate to help them leapfrog into the digital-first age.

A McKinsey and Company paper throws a wrench in the works of many banking technology roadmaps, moving beyond the ‘digital-first’ paradigm and creating a new one, the ‘AI-first’ paradigm. With the maturity of AI technologies, it’s off the races for both traditional and digital financial entities in a new race that will potentially be the race to end all races.

However, before looking at AI use cases in retail banking, any discussion on AI merits a mention of the challenges involved in implementing AI solutions within a finance setting.

Some of the challenges financial institutions face with regard to embracing AI in totality include data privacy and security, ethical considerations especially with regard to decision-making bias and labor replacement, regulatory compliance and cybersecurity risks. Nevertheless, it is imperative that financial institutions embrace AI as a matter of urgency and necessity as downstream users and customers begin the acclimatize to the reality of AI in their daily lives. .

To provide some context on some of these advancements and how retail banking is working AI into core business processes, here are three examples of current and potential future uses of AI in retail banking.

1. AI Virtual Financial Assistants

Almost every bank has implemented some kind of chatbot in its customer facing infrastructure. The idea that armies of customer care representatives are replaceable with virtual chat bots has been a main driver of this trend.

Now, with generative AI like ChatGPT and similar LLMs, chatbots have become increasingly capable of having cohesive and helpful conversations with customers. Generative AI is proving to be the enabling technology that most banking chatbots require to move the next level of effectiveness. Consider a situation where a customer wants information on their credit card balance, but then quickly switches to talking about getting a home equity loan. AI, especially conversational AI built on generative AI models, Expert Systems, and others, are well suited to handle such multi-turn conversations. One example of such a smart virtual assistant is Bank of America’s AI assistant Erica.

According to BOA, Erica is a smart assistant able to assist digital banking customers with bills and payments, inter-account transfers, and other account-related tasks. The bank anticipates that the assistant will cut down customer care requests significantly while also making its digital products more attractive to Millennials. Erica, while advanced, is but the tip of the iceberg when it comes to potential AI applications. Other areas banks will be looking towards on this front include personalized financial advice, real-time financial management, automated investment management, predictive financial analytics, financial literacy and education, and customized financial products.

As the sophistication of recommendation technology advances, its challenges will become more apparent. When AI chat bots start giving complex suggestions for investments or cash management, the acceptance of these recommendations will heavily depend on the users’ trust level. The goal for AI bots is to offer helpful, but often counter-intuitive advice, as that’s the main reason for investing in computing resources. However, people tend to be reluctant to follow advice that they do not understand. Considering the opaque nature of how generative AI functions, the key question is: how can we bridge the gap between comprehending the recommendations made by the bot and actually implementing them?

2. Data-Driven Fraud Detection and Prevention

Fraud is a major challenge for retail banking institutions. According to a recent report from Javelin Strategy and Research, co-sponsored by AARP, approximately 42 million individuals in the United States fell prey to identity fraud in 2021, resulting in a collective financial loss of $52 billion for consumers.

On the consumer side, a 2023 Experian report found that 85% of people expect businesses to respond to fraud concerns. The report went on to find that 51% of consumers considered abandoning a new account opening because of a negative experience. These results illustrate the challenge banks face to balance protecting customer data and providing foolproof access controls while ensuring these measures do not impact the overall customer experience.

Stated another way, balancing negative CX because of fraud concerns could represent a 2x increase in application volume. Clearly, there is enough value available to justify getting it right.

In 2024 and beyond, to combat fraud, almost 60% of companies are either prioritizing or planning to integrate machine learning into their fraud detection and prevention methods. Machine learning assists in analyzing vast amounts of transactions and data, enabling the identification of potential fraud risks.

According to that same Experian report, among those enterprises employing machine learning models, 90% have expressed a strong belief in their ability to detect and prevent fraud, while 87% have a high level of trust in their customer authentication processes. This indicates that the adoption of this advanced technology in fraud prevention strategies is becoming essential for businesses to stay ahead of increasing fraud risks and meet consumer expectations. A question remains: what kinds of investments will banks need to continue to make to stay one step ahead of a clearly ever more sophisticated band of fraudsters who want to compromise this information.

AI is significantly contributing to these high standards through its data-driven approach to detecting and preventing fraud. Banks are harnessing vast amounts of customer data and using AI in conjunction with advanced predictive analytics to construct complex digital profiles of individuals. These profiles, essentially financial behavioral fingerprints, allow for the early detection of unusual account activities, potentially preventing financial losses.

Such AI systems are designed to understand the typical patterns of how and when customers access their accounts, enabling them to flag activities that deviate from these norms. Moreover, AI technologies enhance this protective measure by identifying activities that were previously undetectable, thanks to deep learning capabilities. These AI systems continually adapt and learn, improving their ability to discern what constitutes unusual and potentially fraudulent activity.

3. Integrated Personal Finance Management

The integration of Artificial Intelligence (AI) in personal finance management (PFM) represents a growing trend that holds substantial promise for improving how individuals manage their finances. This innovative approach, exemplified by startups like Wallet.AI and partnerships like that of Clarity Money and Citibank, signals a shift towards more insightful, data-driven financial guidance.

Wallet.AI, a San Francisco-based startup, stands out in this emerging field. Its approach to PFM harnesses AI to assist consumers in making wiser financial decisions, particularly in daily spending. By analyzing data from various sources, including smartphone check-ins and social media updates, Wallet.AI embodies the concept of 'contextual awareness.' This capability allows the tool to provide personalized financial insights by understanding and adapting to individual spending patterns.

This approach mirrors the broader movement of 'Quantified Self' in personal finance. Like health apps such as Weight Watchers and Noom, which aim to modify behavior rather than just monitor diet, Wallet.AI and similar tools use AI to nudge individuals towards more prudent financial decisions. This strategy acknowledges the common human tendency to make suboptimal financial choices and leverages AI to guide users towards better financial health.

The evolution of AI in PFM also highlights the unique position of retail banks in the personal finance ecosystem. Banks, with their extensive customer data and established trust, are ideally placed to implement AI-driven PFM services. Recognizing this opportunity, many are either developing in-house tools or forming partnerships with fintech startups. An exemplary collaboration is between Clarity Money and Citibank, through its venture capital arm, Citi Ventures. Clarity Money offers features similar to Wallet.AI, such as subscription tracking, spending monitoring, and savings tools, actively engaging users in improving their financial health.

This new wave of AI-driven tools, including Wallet.AI, offers a more nuanced approach compared to traditional PFM tools like Mint, QuickBooks, and Turbo Tax. While these established tools focus mainly on budgeting and tax-related issues, AI-driven platforms like Wallet.AI emphasize everyday financial decision-making, providing a more comprehensive solution to personal finance management. The difficulty of the traditional PFM approach is highlighted by the recent decision to close the Mint platform completely as of January 1, 2024. Intuit would like to blend that together with Credit Karma into a “reimaged” offering, but the fact that it cannot continue as is shows its approach wasn’t meeting expectations – neither consumers’ nor investors’.

Positive Steps in the Right Direction

Most banks currently use various types of AI (such as Machine Learning and Expert Systems) in their daily operations and have been doing so for decades. Recently, the discussion around AI has been almost overtaken with talk of generative AI (like ChatGPT) and the next frontier of Large Language Models (LLMs).

A recent dramatic management snafu at AI standard bearer OpenAI thrust the concept of Artificial General Intelligence (AGI) into the mainstay of conversation. AGI is defined by many to be the level of AI that exhibits the kind of cognitive capabilities that humans do.

Current AI is only largely capable of conceptual work, and many retail banks are aware of this limitation. However, the promise of AGI also comes with challenges, the most important being the black-box nature of how it will work, a major challenge especially in circumstances where banks have to explain their actions to regulators. Such ambiguity creates an uncertain environment for banks, which are governed by strict ethical and statutory rules that leave little room for error.

We are starting to see differences in how banks react to the promise of AI depending on the regulatory environment in which they exist. One of the defining characteristics of that environment seems to depend on the balance between the adversarial nature of the regulator versus the regulator’s desire to cooperate with their banking institutions to build the “vision” of its country’s banking system. There is no doubt that this balance differs in Western financial systems from emerging financial systems in Africa, the Middle East, and Asia. How that dynamic ends up affecting these institutions’ customer experience (CX) remains to be seen. Will a more cooperative regulatory environment create a more technologically savvy bank, or will it simply lead to a rickety, unsafe economic system? Time will tell.

However, the steps we are seeing retail banks taking so far, though cautious, are largely positive and provide the groundwork for future, more aggressive, implementations of AI. From a consumer perspective, while most people will not explicitly identify AI as a driver, consumers today want increasingly intelligent and interactive solutions from service providers. As companies like Google, OpenAI, and Microsoft lead the charge on deploying AI services, retail banks must participate in the “dance” to ensure they are not disrupted by these massive tech giants, some of whom would not mind deepening their penetration into the retail banking sector.

Frank Bria is the author of Seven Billion Banks: How a Personalized Banking Experience Will Save the Industry and Managing Principal Consultant at Bria Strategy Group. He is an expert at advanced analytics, machine learning, and artificial intelligence in retail and commercial banking. He has consulted for 8 of the top 10 banks in the United States and many of the top 15 global banks having led analytics technology implementations on 4 continents. His designs and system implementations for credit, collections, pricing, account opening, compliance, risk, and underwriting are still in place at major institutions across the globe.

Frank holds a Masters of Engineering (MEng) in Mechanical Engineering/Systems Engineering from Colorado State University and a Masters of Science (MS) in Mathematics from Purdue University. He is a go-to resource on best practices and technology implementations for financial institutions and vendors alike. He can be contacted via email at: [email protected].

Watch for his new book, Eight Billion Banks: The AI Transformation of Personalized Banking, to hit the market in mid-2024.