Banking Exchange Magazine Logo

Getting up to speed on AI

AI Series: Understanding what “artificial intelligence” encompasses—and why it matters

First of a series: Artificial intelligence promises to be the single-biggest factor in continuing evolution of the banking business. Big banks have been leading the way thus far. First of a series: Artificial intelligence promises to be the single-biggest factor in continuing evolution of the banking business. Big banks have been leading the way thus far.

Machine learning. Deep learning. Robotic process automation. Natural voice recognition (chatbots). These terms—just now coming into the banking vernacular—can be confusing, especially since they are all subsets of the equally dense term artificial intelligence or AI.

It behooves bankers to make the effort to get a handle on what they mean, how they are interrelated, and, most important, what potential they offer to improve customer relationships, reduce fraud, beef up operational efficiency, reduce costs, and, ultimately, add to revenues.

Through a number of interviews with Banking Exchange, bankers, analysts, and a prominent futurist paint a picture of what AI and its various subsets will mean to the banking industry.

“AI is making it possible for customers to engage with companies in more ways than ever before through voice, gesture recognition, video, and chat, while opening up possibilities to serve customers beyond their own digital properties,” says Brad Stewart, senior vice-president, head of product, AI Enterprise Solutions, at Wells Fargo. “We think these will continue to get even more sophisticated. Projects range from systems that can spot payments fraud or misconduct by employees, to technology that can make more personal recommendations on financial products to clients.”

Who is interested in AI?

It’s not just the big banks that are making headway with AI, although that’s where the leading edge of progress is.

“We are just in the first couple of innings of this,” says Michael Abbott, managing director of financial services/digital, North America, at Accenture. “The transformation is going to occur over the next four to five years. It is going to be spectacular.”

Peter Graves, chief information officer, Independent Bank Corp., Grand Rapids, Mich., says his bank uses a form of AI mainly for fraud mitigation, but sees the day when it could be more widely applied. The immediate hurdle is to coach the bank’s existing customers—many of whom require personal interaction—to accept a digital interaction that is indistinguishable from that of a human.

“Getting them [existing customers] to adopt it is a challenge,” Graves says. “It has to be smooth and personable as much as possible. Otherwise, you feel like you’re still talking to a machine on the other end of the phone.”

The goal, says Graves, is “to make it personable even though it’s digital. That enables behavior transition to happen much more quickly.”

Chris Nichols, chief strategy officer at CenterState Bank, Winter Haven, Fla., makes no bones about it.

“I consider AI . . . as probably the single biggest factor that will change banking,” Nichols says. “It has widespread applications across many areas of the bank.” This includes fraud prevention, marketing, branch selection, profitability, and underwriting applications at his bank.

Where does AI stand?

Still, AI now is in the realm of large banking organizations with lots of resources to apply to these particular technologies as well as simultaneously to all the rest of the technological upheavals affecting the banking industry.

Here are a couple of anecdotal examples that are gleaned from the internet:

• The Royal Bank of Canada recently hired a PhD pioneer in AI as head academic advisor to RBC Research for machine learning studies.

• Lloyds Banking Group, London, recently placed an employment solicitation for someone to join its information management group—specifically to lead its machine learning team.

Looked at in a more quantifiable way, however, a couple of surveys illustrate why actual implementation of AI today stands at a very early stage in banking. On the one hand, Accenture recently found that 82% of the U.S. bankers it polled believe that AI will revolutionize the way banks gather information and interact with customers, and 72% believe that within three years banks will deploy AI as their primary method for interacting with customers.

On the other hand, in December, Celent issued a report in which it polled 100 U.S. bankers and asked them which emerging technologies were most important to deliver their top priorities. On the top of the list was mobile banking channel development, at 96%. Last on the list was AI-based initiatives, at 6%.

“Our best hypothesis [for the low showing of AI] is that it’s because the smaller banks in particular, below $50 billion in assets, are so overwhelmed and so preoccupied with all this other stuff on their to-do list,” says Daniel Latimore, senior vice-president, banking, at Celent. “AI is something they are keeping their eye on and they would like to do more in, but they have other stuff to do that requires some degree of expertise.” Nevertheless, he points out, while most banks will be followers on AI, “they should definitely be paying attention to it.”

What is AI?

So what is AI, and how is it related to these other terms: machine learning and robotic process automation (sometimes shortened to robotics or RPA)?

“At Wells Fargo, we define artificial intelligence as a global term encompassing a number of individual and nested constituent technologies, disciplines, and scientific fields,” says Stewart.

In a similar vein, Antonis Papatsaras, chief technology officer at SpringCM, an enterprise content management company based in Chicago, explains it this way: “Think of a concentric circle where the outer circle is artificial intelligence. The next inner circle is machine learning . . . Then, in the center, is robotic process automation.”

To be sure, there are subsets of subsets.

In machine learning, for example, there is deep learning, which generally supersizes the amount and sources of data with which the machine learning technology absorbs and uses. In addition, there are AI-related technologies that blend, such as natural voice recognition, which can work with both machine learning and robotics, as in the use of chatbots—systems that can intelligently converse with customers.

In general, then, AI is composed of “technologies that mimic human judgment at high speed, high scale, and low cost,” says Sridhar Rajan, robotics and cognitive automation lead for Financial Services, Deloitte Consultants.

Given that AI is the encompassing term for a number of related technologies, it is important to distinguish them.

What is machine learning?

Wells Fargo’s Stewart defines machine learning as “a branch of AI that utilizes data and algorithms to train software logic instead of programming that logic through explicit rules.”

CenterState’s Nichols puts it this way: “When we talk about machine learning, we basically talk about using a set of algorithms that gets smart, that improves the base algorithm. It draws conclusions from it and learns over time.”

Thus, in a simplified example, a given machine learning algorithm may dive into a vast sea of data and draw correlations between various, apparently disconnected inputs—transaction history, social media postings, customer locations, device types—and decide whether or not a particular transaction is fraudulent or not.

That decision may or may not be correct. If it’s not correct in a given instance, a human operator can step in and make an adjustment to the algorithm. Afterward, should similar instances arise, the machine will have learned not to make the wrong decision again.

Deep learning is a subset of machine learning and basically funnels massive amounts of data from many different sources very quickly into a neural network (modeled after the human brain), and then responds to specific questions about specific customers or trends.

Robotic process automation

Moving to another subset, RPA, SpringCM’s Papatsaras points out that it involves “tools that automate a special work environment, taking tasks that humans are bogged down with . . . to achieve consistency.”

In other words, in a rules-based system in which every time X, Y, and Z are considered, the outcome must be A (an “if-then” circumstance, says Latimore). The RPA system can do that task better and faster than humans.

“If you get humans involved, the human emotion, human fatigue, and mentality of the moment could affect the outcome of the [task],” Papatsaras says, even though the outcome should always be the same given all the factors involved. With RPA, he points out, every customer can be assured of having been treated fairly and consistently, which then increases customer satisfaction.

Good data is required

Crucial to all AI applications is their dependence on good data—the more the better. “Banks are built upon data,” says Chris Skinner, financial technology futurist and noted author on the subject, including the recent book ValueWeb. “That data is what will power these machines and these robotics and these intelligence capabilities.” 

Adds Papatsaras: “It’s about a huge, insane amount of data that we collect these days in a bank organization—about our customers and their behaviors in our system, about advanced algorithms that parse that data and make a prediction or a determination.”

The whole point of AI, in fact, is to quickly and seamlessly assess the disparate sources and immense quantity of data in order to, as Independent Bank’s Graves says, “bridge the gap between the information and the end user.”

So who is most likely to have the “huge, insane amount of data” for which AI likely is most useful? The easy answer is: the largest banks.

“There is a richness of data, an abundance of data in large financial institutions,” says Sasi Mudigonda, chief product strategist for financial crime and compliance at Oracle Financial Services. “I’m not saying for smaller financial institutions it doesn’t exist, especially if they pool the data over time.”

Graves sees a potential avenue to pursue for smaller banks, and that is the use of aggregator services that can collect all of a given customer’s banking relationships—not just within one bank—and put all that into one place.

“What it does is it creates a database that is bigger than what you have in [the relationships with one bank],” says Graves. “If you had five to ten relationships with a bank customer, you think you’re doing pretty good. But what if you had 20 to 30 relationships that are all put in one place where the data from that is now bigger than just what’s in your own bank?”

AI improves relationships

When it comes down to it, the crucial potential of AI and its manifestations is reflected in how it can substantially improve customer relationships.

“We’ve identified use cases for AI that can ultimately help in enhancing customer experience, protecting against fraud, and furthering our application of deep-learning algorithms,” says Wells Fargo’s Stewart.

With AI, and particularly its application to chatbots, says Graves, customer interactions can combine with actual savings to the bank. “You [can] apply the AI around voice recognition and big data, then you can start to have a conversation with the customer that is completely digital, but that has some meaning to the consumer and gets them the information they want. . . . You don’t have people sitting around doing this—that’s the huge savings.”

What’s to come

Looking forward, Celent’s Latimore sums up the likely progression of benefits as AI seeps into the banking industry in the coming years: “Decreasing expenses is probably the No. 1 rationale. Mitigating risk is No. 2, on the fraud side. Then the revenue side is No. 3, for now, but they all play in all these things.”

The futurist Skinner puts it another way: “The banks that get the best intelligence out of their data will probably get the majority of the market in the next decade.”

This is Part 1 of a three-part series. This article originally appeared in the June-July edition of Banking Exchange magazine.

Part 2: What you need to know about robotics

Part 3: What you need to know about machine learning

John Ginovsky

John Ginovsky is a contributing editor of Banking Exchange and editor of the publication’s Tech Exchange e-newsletter. For more than two decades he’s written about the commercial banking industry, specializing in its technological side and how it relates to the actual business of banking. In addition to his weekly blogs—"Making Sense of It All"—he contributes fresh, original stories to each Tech Exchange issue based on personal interviews or exclusive contributed pieces. He previously was senior editor for Community Banker magazine (which merged into ABA Banking Journal) and for ABA Banking Journal and was managing editor and staff reporter for ABA’s Bankers News. Email him at [email protected]

back to top


About Us

Connect With Us