by Bruce Harpham

Banks turn to AI for improved customer service, competitive edge

Feature
Sep 22, 2016
Emerging TechnologyEnterprise Applications

2 online banking
Credit: Thinkstock

With the Wells Fargo fake accounts scandal fresh on your mind, it’s hard to feel sorry for banks these days. But the pressure on big banks to innovate has never been higher.

The heady days of the 2000s, with their high profits from lending, booming investment banking and relatively light regulation, are gone. New laws and regulations have forced banks to allocate large budgets to regulatory compliance. At the same time, payment services like PayPal and investment services like Wealthfront are competing for the clients that banks used to take for granted.

Artificial intelligence may prove to be a key part of the solution for banks seeking to remain competitive. For an industry defined by personal relationships, a cultural change is underway as technological customer service and traditional customer service are increasingly serving different segments. High net worth individuals continue to be served by professionals, while simple banking products — credit cards, loans, checking accounts — are being pushed toward self-serve options and now AI.

Here’s a look at what three banks — Capital One, RBS and TD Bank — have learned from adopting AI.

With the Wells Fargo fake accounts scandal fresh on your mind, it’s hard to feel sorry for banks these days. But the pressure on big banks to innovate has never been higher.

The heady days of the 2000s, with their high profits from lending, booming investment banking and relatively light regulation, are gone. New laws and regulations have forced banks to allocate large budgets to regulatory compliance. At the same time, payment services like PayPal and investment services like Wealthfront are competing for the clients that banks used to take for granted.

Artificial intelligence may prove to be a key part of the solution for banks seeking to remain competitive. For an industry defined by personal relationships, a cultural change is underway as technological customer service and traditional customer service are increasingly serving different segments. High net worth individuals continue to be served by professionals, while simple banking products — credit cards, loans, checking accounts — are being pushed toward self-serve options and now AI.

Here’s a look at what three banks — Capital One, RBS and TD Bank — have learned from adopting AI.

Capital One’s voice bot

In March 2016, Capital One launched a new chatbot service in partnership with Amazon. “We were the first bank to launch on Amazon’s Echo platform. We’re pleased with the volume of activity through the service so far,” says Ken Dodelin, vice president of digital product development at Capital One.

The Alexa platform is a voice-activated service from Amazon that combines hardware (e.g., the Amazon Echo device) and software (the Alexa Voice Service). Currently, third-party companies on the platform include Honeywell, Uber, 1-800 Flowers and Domino’s Pizza. Unlike web or smartphone services, the Amazon service requires the purchase of hardware, which may slow adoption, but Fast Company’s Mark Sullivan thinks Alexa is poised to “become the go-to AI for the home.” For companies, Amazon smooths the development process by handling the speech recognition and related technical challenges.

Capital One’s Amazon-based virtual assistant provides a number of services to customers. “The Capital One Alexa skill started with support for checking accounts and credit cards. In July 2016, we added support for home loans and auto loans,” Dodelin says. The development work, which was done in-house at Capital One with periodic feedback from Amazon, took approximately seven months from beginning to launch.

“Unlike mobile apps, updates and new features can be added automatically without the user having to download an update,” Dodelin adds. “Based on our user research, the highest priority is for customers to check on their own accounts rather than asking general questions about products.” Popular questions include “When is my balance due?” and “What’s the balance in my account?”

Defining a narrow set of use cases was key to Capital One’s success. “Our focus was on simple, repeatable use cases,” Dodelin says. “Learning how to write scripts that sound appropriate was a key part of the learning process for us. When we first wrote a script and tested it, it often sounded quite different from what we imagined, so we would rewrite that section,” he said.

As with any new technology, there have been some unexpected results. “I have heard of a few cases where customers have asked Alexa to pay their credit card balance instead of executing a payment,” Dodelin says.

RBS deploys AI to support staff

Most approaches to AI emphasize how the technology supports end customers. The Royal Bank of Scotland (RBS) took a different approach. Launched in March 2016, the Luvo service, which is based on IBM’s Watson technology, helps RBS employees respond to customers more quickly. Offered as a web chat service, Luvo is focused on customer service inquiries such as “My customer has lost their card – what steps do they need to take now?” and “My customer has locked their PIN – how do they unlock it?”

The RBS experience presents several lessons for banks and other companies to consider for their implementations:

  • Like Capital One, RBS opted to partner with a technology leader for the core AI capability rather than building it from scratch.
  • The company took inspiration from outside its industry — WhatsApp, the popular chat service, has been cited by RBS leaders as a key source of inspiration.
  • Luvo was launched as a pilot to about 1,000 staff to evaluate performance before considering next steps.
  • The company seeks to adapt the service to customer inquiries rather than guide customers into pre-programmed queries.

TD Bank adds AI to apps

Unlike RBS and Capital One, TD Bank’s approach is based on adding a new AI capability to its existing apps for iOS and Android. “TD Bank worked with us to add ‘context as a service’ to their smartphone app,” explains Hossein Rahnama, founder of Toronto-based Flybits. Launched in the summer of 2016, the AI-powered feature already has one million users.

“Mortgage renewal is a key priority for the bank and our app addresses that. When a TD bank customer with our app walks past a branch and their mortgage is close to renewing, they are invited into the branch,” Rahnama says. The app also offers location-based direct response marketing. TD credit card customers traveling through Toronto Pearson International Airport are offered travel upgrades.

Delivering reliable AI to bank customers draws on several disciplines. “I suggest starting small with a single use case when building this kind of capability. It’s also vital to pay attention to design and user experience — there needs to be a clear sequence and no overwhelming menus,” Rahnama said. Privacy and security are paramount concerns.

“Our approach is to educate customers on privacy and use an opt-in approach. Our app also includes a privacy manager where customers can easily change their settings,” Rahnama says. There’s a fine line between pleasantly surprising customers by anticipating their needs and knowing too much.