A few years ago, when the use of artificial intelligence-enabled chatbots was becoming widespread across multiple industries, particularly in customer services operations, New York-based consultant Adelyn Zhou expressed the question being asked by many tech vendors trying to sell their products to the financial services industry: “So why are banks – who are typically the most capable and tech-intensive players in the business world – acting like Luddites with AI?”

In the financial services industry, balancing the potential risks and rewards of adopting new technology has always been a tricky business. There are many reasons why banks move slowly in some areas, but one of the biggest is concern about damaging customer trust. In an industry where a reputation for solidity and prudence are at a premium, the possibility that unforeseen consequences of new technology could have embarrassing, or catastrophic effects on customer relationships has naturally led to caution.

Reputational concerns, in part, explain the relatively late adoption of artificial intelligence-enabled voice recognition systems in the industry.

It’s true that the iPhone’s now ubiquitous personal assistant “Siri” has been around in some form since 2011, and Amazon’s voice-activated “Alexa” speakers have been taking customer orders since 2014. But anyone who was paying attention to these kinds of systems in their early incarnations remembers more than a few very public hiccups. They were initially unreliable, frustrating, and often poorly designed, leading to frequent public ridicule.

In 2011, the British Broadcasting Corporation released a comedy sketch that went viral, depicting two Scotsmen trapped in a voice activated elevator that was unable to understand their accents. As recently as 2016, Microsoft was forced to disable an AI chatbot linked to a Twitter account less than 24 hours after it launched, because users taught it to deliver racist, sexist, and obscene replies.

For smartphone makers and even retailers, the bumps in the road on the way to efficient AI chatbots may have been embarrassing, but they weren’t fatal. For financial institutions, though, the risks are far greater. A malfunctioning chatbot that you use to order cat food isn’t nearly as worrisome to a consumer as one connected to the financial institution that holds your life’s savings.

However, it has become obvious that banks can no longer afford to wait to implement AI-enabled customer service systems. And the good news is that at this point, the technology is sufficiently advanced that they don’t need to.

In a paper last year, Forrester Research warned that automated voice and chat interactions are an “increasingly critical functionality” across multiple industries.)

“Voice and chat interactions are quickly moving into the mainstream,” the authors of the study found. “The ability to talk to devices at home; request information through a chat interface; and place an order while driving a car, using only your voice — as well as a host of other use cases — are moving from the realm of science fiction into daily life.”

This is nowhere more true than in the world of financial services, as experts from Accenture pointed out in recently published research. “Financial services companies, too, are entering the intelligence age,” they write. “And they’re doing so while already under intense pressure on multiple fronts. Rapid advances in AI are coming at a time of widespread technological and digital disruption. Competition is fierce.”

Implementation of AI-enabled chatbots may be necessary for banks that want to remain competitive in a changing marketplace, but that doesn’t mean it will be simple. According to a PWC analysis, the typical architecture of a chatbot is made up of four major elements, all of which are relatively complex in and of themselves.

At the base of the system are its record sources. These are sources of structured and unstructured data, including the institution’s own records and publicly available information extending to social media accounts, news archives, government data, survey data, and more.

Sitting on top of the record sources is a platform for what’s called a “data lake.” This consists of a data consumption interface that allows the system to pull information from its various data sources, a data storage element that makes the information accessible to a data processing system, and finally an insight delivery layer that delivers data to the AI platform.

The AI platform itself is made up of multiple layers tailored to the specific way in which the data will be delivered, typically including a web layer and app layer, which draw information from an internal database.

Finally, comes the public-facing customer interface, which can take the form of a website, a mobile application, or a phone-based system, all of which must connect seamlessly to the underlying AI platform.

For those able to adapt the new technology successfully, Accenture’s analysts add, “Artificial intelligence will enable financial services companies to completely redefine how they work, how they create innovative products and services, and how they transform customer experiences.

Banks Using AI See Rewards

Indeed, for banks that are on the cutting edge of implementing these new tools, the rewards are very obvious.

Regions Bank, the Birmingham-based institution with branches throughout the South and Midwest, has implemented a pair of AI-enabled systems to help it manage customer inquiries. “Reggie,” a system built on IBM’s Watson platform, answers customer calls, authenticates them, and passes them on to the proper channel within the bank. According to Regions officials, the system paid for itself within a year by cutting average customer hold times at the bank’s contact center by nearly 30 seconds.

Regions’ second, more complex AI system is known as “Rosie.” An amalgam of nearly two dozen data models that pull product information and customer data from systems throughout the bank, Rosie is designed to produce what developers call a “next best action” decision.

In a 2018 presentation, Allison Nygaard, a Regions senior vice president for business transformation, explained that Rosie can interact directly with the Reggie system, giving the chatbot a recommended response to a customer query. However, the system can also be used by human bankers in real-time interactions with customers to help identify clients needs or potential opportunities to market new products.

Rosie, a bank executive told the publication The Financial Brand, generated an internal rate of return of nearly 50 percent, three times that of most of the bank’s tech products.

The combination of a highly functional chatbot with a powerful AI application able to process vast amounts of structured and unstructured data offers extraordinary promise for the banking industry.

Bank of America, an early adopter in the industry, launched its voice activated assistant, Erica, in 2016. In March, it announced a suite of enhancements to the product, and reported that in under three years it had amassed more than six million users, answered more than 35 million client requests, and learned upwards of 400,000 different ways in which questions can be asked.

How Do Chatbots Benefit Banks?

The service’s most popular queries are transaction searches that allow users to specify certain characteristics of a transaction — merchant, amount, time frame, etc. — and returns a list of all hits going back 18 months. It also allows users to make real time queries about their spending habits, check their credit scores, and more.

In some ways, it’s the ability to help customers with the smallest of things that offers the most benefit, Christian Kitchell, head of artificial intelligence and Erica at the bank told American Banker newspaper. One of the most frequent calls to the bank’s support center come from customers who simply need the bank’s routing number.

“It’s the kind of thing you don’t need every day, so it’s hard to know where to look for it in the app, so Erica is a great way to get that in with a simple utterance,” he told the paper.

While banks may have been late to the game, relative to other industries, in adding AI-enabled systems to their customer service arsenals, the competition to implement and leverage these systems is heating up quickly.

In a warning to the industry, Accenture’s Jubraj, Graham, and Ryan predict problems for institutions that fail to keep up.

“As high-powered computing becomes ever more readily available, and as vast data sets needed for training AI solutions become more accessible, the capabilities will continue to grow exponentially,” they write. “The world has barely scratched the surface of AI’s possibilities. That creates an imperative for all financial services companies. The time to move on AI is now. Low barriers to entry will bring ever fiercer competition for AI talent, AI patents and AI capabilities. And coming AI advances will be so all-encompassing, and so fast-paced, that high-performing organizations will inevitably accelerate away and leave the slow movers far behind.”

Further Reading: