Close this search box.

Is it Time To Go All In on AI in Banking?

Over the past couple of decades, the banking industry’s view of artificial intelligence and machine learning has evolved substantially. AI has progressed from a pie-in-the-sky promise to a trendy buzzword to its current status as an influential and growing factor in day-to-day operations. 

The appeal of AI is obvious: In a 2019 survey by McKinsey, 63% of respondents reported revenue increases in business units where AI had been adopted, and 44% reported cost savings. 

That said, the prospect of improved revenue and reduced costs is alluring. To break down the potential effects of AI in banking, let’s look at several applications and how they might affect your institution. 

Leveraging AI for Growth and Improved User Experience

Let’s begin with AI’s potential in customer-facing roles. When your marketing team speaks of the need for an updated digital platform, this is usually what they’re talking about: The rest of your business plan stands or falls on your ability to attract and retain users. And AI has tremendous potential to help you do that. 

Areas, where AI can revolutionize your user experience, include: 

World-Class Personalization

One of the key reasons for fintechs’ growth (and similarly, a reason the tech giants will become a serious threat as they edge into the financial marketplace) is their advantage in personalization. They’re simply better at giving customers what they want when they want it. Gaining that capability for your own institution takes a key competitive advantage away from fintech, and — when executed well — can drive user loyalty

Improved Analytics

This is why your rivals may be better at personalization. AI can help you navigate the oceans of data your users generate, parsing it into specific needs and matching those to your corresponding products and services. 

More Targeted Marketing

Your users will view the outcome of improved analytics as top-tier personalization (“they really get me”); but the flip side of that coin is more targeted, effective, and cost-efficient marketing. When your analytics can drive focused, in-the-moment offers at the single-user level, designed to arrive exactly when they’re needed, you’ll have arrived at a very sweet spot in your marketing plans

Improved Customer Service

AI can do a lot of things on the customer-service side of your operation, from mundane efficiencies (calculating staffing requirements, routing phone calls) to more delicate tasks such as recognizing natural-language searches on your site and operating interactive chatbots or voice assistants (like Siri and Alexa, or Bank of America’s Erica).  

Don’t Sleep On the Importance of Customer Service

Improved customer service may not bring as much immediate appeal to C-suite strategy sessions as those other bullet points, but Evan Siegel sees it as a significant use case for AI in banking. A 15-year veteran of strategy and customer-experience roles at Wells Fargo, he’s now the VP of financial-services AI at eGain Corporation. 

“I liken customer service to a triangle,” he says, “the familiar ‘hierarchy of needs.’ At the bottom of the pyramid, you have the basic transactional competencies of things being done correctly the first time, more often than not. The next level is convenience (e.g., mobile deposit), including where if something does go wrong, it’s corrected quickly and without inconvenience to your user. 

“The third level goes beyond that, to helping users realize their financial goals. It’s something that shows up in the mission statement of almost every financial institution, but few actually do it well. This is a great use case for AI, because AI makes it possible to provide personalized financial guidance at scale. This is important, because what’s good for the customer is good for the financial institution as well. Consumers who are realizing their goals have more money, build better credit and are able to take advantage of a wider range of products and services. At the end of the day, they generate more business.”

Siegel singles out coaching and financial advice as an excellent AI-driven “value add” for financial institutions, pointing to a pair of his own products as examples. “We have a solution that guides team members through a needs assessment when they’re sitting down with a client,” he says. “The system prompts the team member to suggest the next best product or financial action, and creates a record of what was recommended and why. It’s all auditable, with a record of customer answers and CRM data used to make recommendations.

“That’s a great in-branch application of AI, but we also have a product that provides completely AI-driven financial coaching 24/7. This makes high-quality financial guidance universally available, instead of confining it to the relatively small subset of high-net-value clients.”

AI in Underwriting and Risk Management

Attracting potential business to your institution is one thing; ensuring that the resulting business is worth writing is entirely another. This is a second area where AI and machine learning hold a great deal of potential. The rise of the gig economy and precarious, multisource employment was already a challenge for the industry, and the disruptions caused by the COVID pandemic have made it ever harder to assess who is and is not creditworthy. 

AI provides an immeasurably powerful tool for assessing risk, able to assess volumes of data that would make a human’s head spin and tease out the meaningful relationships within them. That said, it’s a tool that requires a substantial degree of testing and oversight. If done correctly, it can eliminate the biases that have left some communities underserved, but if done poorly it can perpetuate or even exacerbate them.

Verifying the models you use can be the difference between getting the result you want and accepting an unaccountable “black box” that spits out results you can’t verify or explain. In one case, researchers trained a machine-learning algorithm to distinguish between images of huskies and wolves. Looking deeper into a missed identification, researchers realized that all the wolf photos they’d used for training purposes had snow in them. They had inadvertently trained their algorithm to consider snow a “wolf” characteristic.

It makes an amusing story for data scientists to tell over coffee or while warming up the audience at a presentation, but it’s decidedly less so for financial institutions who could miss out on business if their models are inaccurate, or potentially face litigation if the models show demonstrable bias.  

AI in Collections and Recovery

The flip side of underwriting and risk assessment is collections and recovery. AI and machine learning don’t have a direct role to play here, but their indirect role can be substantial. 

For any financial institution, it’s better and less costly to keep accounts out of collection in the first place. The same AI tools that drive your personalization and marketing efforts can be made equally adept at identifying clients whose finances are trending in the wrong direction. Sometimes it will be obvious — the sudden disappearance of employment income, or its replacement by unemployment benefits or insurance payments; other times it may be more subtle, like a steady increase in debt or a pattern of later bill payments. 

AI can detect these patterns and intercede while there’s still time to make a difference. This might take the form of an in-context warning offered by your app, or a link to a pertinent video on your site, or even in-person coaching from one of your staff. In scenarios where a crisis has arrived abruptly (unemployment or injury, perhaps), AI can flag the user’s account for direct contact. Reaching out and proactively offering to help restructure debt or payments, or even pause them for a time, can help preserve your client’s creditworthiness (and build loyalty toward your institution). 

AI in Managing and Streamlining Internal Processes

On the “reduce costs” side of the ledger, one of the most promising aspects of AI and machine learning is the potential to manage and streamline an institution’s internal processes. Such systems can take over repetitive activities from human staff, for example, freeing them up for more productive uses of their time (and being freed of drudgery can be a great morale booster for your frontline team members). 

That effort won’t necessarily be restricted to the activities of your lower-level employees. For example, JPMorgan Chase rolled out AI-driven “contract intelligence” software in 2017, using it to analyze the contracts governing its tens of thousands of commercial credit agreements. Reviewing the contracts manually took hundreds of thousands of hours; the software did it in seconds.  

Your IT department is another area where skilled employees spend much of their time on repetitive tasks, and AI can automate many of those. For example, algorithms can be used to test code ongoingly throughout the development process, identifying bugs well before they reach the production stage. Mission-critical code still requires a final audit from skilled human eyes (Lumin Digital uses a process like this in-house), but working this way can largely eliminate the need for regression analysis at the time of deployment. 

Managing the regulatory environment, to help institutions remain in compliance with everything from their fiduciary duties (as Siegel pointed out in a recent article for the Filene Research Institute, “bots don’t go off-script”) to local employment standards to federal regulations and money-laundering laws, is another important application of AI. Numerous banks have paid billion-dollar fees for violations of U.S. sanctions, for example, and AI tools can manage this kind of complex analysis in a timely and effective fashion.  

AI in Automating and Improving Threat Management

One final — and extremely important — use case for AI is in the area of threat management. Small and midsize financial institutions lack the resources of their larger peers, and as a result can be especially vulnerable to malicious actors. From individual users falling victim to identity theft to large-scale data breaches and ransomware attacks, the range and scope of potential attacks is intimidating. 

AI can help thwart those vulnerabilities, by identifying and flagging activities that are symptomatic of attacks. This is a more powerful and flexible approach than many of the traditional defences, such as blocking specific IP addresses. It’s analogous to the revolution that occurred in antivirus software over the past couple of decades: Early programs checked for known pieces of malicious code, known as virus signatures. Modern programs rely more heavily on cloud-based AI to recognize malicious activity in real time, regardless of the underlying code (“if it walks like a duck and quacks like a duck…”), and halt it before it actually reaches the user’s computer. 

Similarly, AI in banking can detect unusual activity on a user’s account and flag it for scrutiny as a potential account takeover, or detect attempts at privilege escalation and other staples of the hacker’s toolkit. Your institution may not have the in-house IT skills or flexibility to keep up with the rapidly evolving ecosystem of cybercriminals, but AI can remedy that deficiency. 

AI in Banking as Part of the Big Picture

Making the best use of AI at your institution is a challenge, in large part because it has traditionally been approached on a reactive, scattershot basis: It’s been reactively deployed as a solution to a specific issue or opportunity. In a recent paper, Deloitte argues that AI must instead become an organic part of your institution’s larger operational strategy. 

In Deloitte’s analysis, success (defined as sustainable outcomes) requires AI to be scaled for deployment across the enterprise. That requires a substantial investment in skills, personnel and software development, which may be out of reach for many (or most) small to midsize institutions. That’s where Lumin Digital can help. 

Our digital-banking platform is built on a base of powerful, cloud-native code with AI integrated throughout, effectively placing your institution at the end stage of Deloitte’s road map. You’ll also be able to add new capabilities, including further AI tools, through our powerful set of Application Programming Interfaces (APIs). 

Contact us today for a demonstration, and the opportunity to learn how Lumin Digital can help take your institution from laggard to leader in the AI race. 


McKinsey & Co: Global AI Survey: AI Proves its Worth, but Few Scale Impact

Future Digital Finance: One Million People Are Now Using Erica – BofA’s AI-Powered Chatbot

Brookings Institution: Reducing Bias in AI-Based Financial Services

Harvard Business Review: AI Can Make Bank Loans More Fair

Filene Research Institute: Machine Learning: What? Why? And the Huh???

Emerj: AI in Banking – An Analysis of America’s 7 Top Banks

Financial Times: Banks Adopt AI to Manage Sanctions and Compliance Risk

Filene Research Institute: Humanizing the Digital Experience: How Virtual Financial Coaching Can Help Credit Unions Win Again on Member Service