When Capital One rolled out its automated fraud-check texting system a few years ago, the bank was surprised to find how many possible answers its customers could find to a simple yes-or-no question.
While around 85 percent of people responded to the alerts about potentially suspicious credit card activity with a “confirm” or “deny” as directed, the rest of the replies ranged from fat-fingered typos to a “yep, that was me” or even intimate bits of travelogue—“yes, that’s the purse I bought in Philadelphia last weekend while visiting my sister”—says Ken Dodelin, the bank’s vice president of conversational artificial intelligence product development.
“I like to say that we launched the chatbot a long time ago—just, nobody ever told us,” he says. “Because people were chatting away with us, and we were on the other side not really prepared.”
Capital One wasn’t the only one feeling pressure to improve its technology. The retail banking industry has been increasingly embracing digital automation to keep its lead against financial tech upstarts like digital lenders SoFi and GreenSky—companies that understand how machine learning and personalization can improve customer experience.
Significant investment and institutional overhaul have been powering this move toward automation: Banks have been spending heavily on AI talent, operational upgrades and strategic tech partnerships. One recent report from the International Data Corp. found that the banking industry is the second biggest investor in AI technology worldwide behind retail. It’s expected to sink a total of $5.6 billion into applications like automated customer service agents and fraud prevention in the coming year.
As breakthroughs in AI research make software that can interpret varied voice and text commands, predict user actions and personalize messaging at scale, the industry is latching onto the opportunity to simultaneously cut costs and expand the functionality of one of its biggest expenses: customer service.
“The financial industry is uniquely ready to make a move in the space and bring that computational experience,” says Jason Mars, CEO of Clinc, a startup that makes chatbots for financial institutions like Barclays and USAA. “They’re trying to bring computational AI into the call center to solve some of their biggest pain points—the pain points that have to do with providing customer care and staying relevant in an industry where folks are now moving to this new user interface.”
Banks take the lead
The banking industry is by no means alone in chasing the cost-cutting dreams that the current boom in analytic-AI technology promises. From hyper-personalized marketing messages to leaner supply chains, companies of all stripes are scrambling to scale their data and analytics tools across all divisions in the face of Silicon Valley competition.
But what is different from categories like retail and media is that the incumbent banking industry actually seems to be winning. Rather than bowing to fin-tech startups, big financial institutions have used a combination of savvy collaboration, acquisitions and a willingness to shell out for high-demand talent to stay on top or partner with newer rivals. Some reports also credit the banking industry’s mastery of regulatory compliance issues, which are more restrictive than those of other industries and therefore make banks harder to disrupt.
In Capital One’s case, it took the bank some time to figure out how to respond to those unexpected communications from its customers. The company spent the ensuing months becoming more like a tech company—marshaling engineering talent, upgrading back-end infrastructure and building out AI capabilities like machine learning.
One of the most prominent results of that process is Eno (one spelled backward), Capital One’s genderless text-based assistant, which uses a form of AI called natural language processing to capably understand a range of normal human communication patterns.
“We had this fascinating evolution of training the machines to understand the humans rather than the humans to understand the machines,” Dodelin says. “With conversational AI, [customers] can just express whatever’s on their mind either through voice or text. … It introduces incredible complexity and variability into that interface.”
Eno was the first chatbot of its kind from a U.S. bank, but it was soon joined by contemporaries across the industry: Erica at Bank of America, the Royal Bank of Scotland’s Cora and Marge, and Bank of Montreal’s Bolt, among others.
These virtual assistants are expected to trim hundreds of millions of dollars in service expenses as the technology becomes one of the dominant forms of addressing customer issues. A study from Juniper Research in February forecast that industrywide cost savings from chatbots could reach $7.3 billion by 2023, a 3,400 percent increase from the estimated $209 million they are expected to provide this year. That sum would represent a reduction in call-center time equivalent to 862 million hours, or nearly half a million working years, according to the firm.
“Chatbots in banking allow heavily automated customer service, in a highly scalable way,” says the Juniper report’s author, Nick Maynard, in a statement. “This type of deployment can be crucial in digital transformation, allowing established banks to better compete with challenger banks.”
Making the technology work for them
That conversational-AI-centric future wasn’t always bright. At the same time that banks were struggling to gain a foothold in tech, chatbots went from being hailed as the new frontier of marketing and customer service across the consumer-facing landscape to what the research firm Gartner terms the “trough of disillusionment” (a phrase from its chart describing the typical hype cycle around emerging technology).
An initial batch of branded-messaging tools brought forth in part by a big push from Facebook Messenger in 2015—the “peak of inflated expectations,” per Gartner’s trajectory—promised universal functionality but fell short with restrictive response options. Like Capital One’s first stabs at texting alerts, they were often barely more sophisticated than age-old automated phone calls. At the time, an internal Facebook study leaked to The Information found that only 30 percent of the bots could fulfill tasks without human help, and the social network wound down the program.
But many banks may have since found a way to move chatbots onto Gartner’s “path to productivity”: Narrow the scope to a single customer-service function, then train the machine behind it with enough data and natural-language-processing ability to make it versatile in fulfilling that need, industry leaders and analysts say. It’s a strategy well-suited to the myriad complex but ultimately rote customer transactions that cost banks big money.
“Banks have come to realize AI is not ready to work miracles yet,” notes Forrester analyst Aurélie L’Hostis, who specializes in digital strategy for retail banks. “You really need to have a specific approach to drive your strategy. You can’t just launch a chatbot without knowing why you’re doing it.”
Making strategic partnerships
Though banks have been investing heavily in developing their own solutions, they’re also partnering with outside firms to keep their edge.
“Most tech in banks has been very old, particularly behind the scenes. And as a result, it’s often taken banks a bit of time to adapt to new technology,” says Jake Tyler, co-founder and CEO of Finn AI, which makes chatbots for big financial institutions like the Bank of Montreal. “But what you have now is a more competitive market for customer experiences thanks to nonbank players entering. … And as a result, there’s a lot more pressure on banks to do more digitally.”
Even a company like Capital One, which has built its own tech foundation, will tap third-party firms to handle certain gaps in its AI abilities, says Dodelin.
“For larger banks, there has historically been a tendency to build in-house,” Tyler says. “As a market, we’re sort of seeing that change at a macro level, to banks doing a better job of working and interacting with a vendor landscape—particularly smaller vendors.”
Beyond the chatbot
Banks are already looking beyond just the automation of existing customer service operations. Many institutions see personal banking eventually evolving to play more of a money-management role in customers’ lives, made possible by personalized data and predictive behavior algorithms.
Ankit Bhatt, svp of omnichannel experience at U.S. Bank, says the developers behind the new AI-integrated mobile app the company debuted earlier this month looked to tech companies like Uber and Fitbit for inspiration in seamless, customer-friendly interfaces. While this first iteration will focus on a range of predictive capabilities, the app is built to eventually accommodate a virtual assistant and personalized content delivery with financial literacy tips and other budgeting advice.
“One of the important findings from our customers is that they really expect the banks to play a pretty critical role to help them become more informed and educated about their day-to-day finances,” Bhatt says. “Our goal is to really improve our customers’ lives—and really become central to their lives.”