Back office, or out front? Bankers differ on where to use AI
Most banks view artificial intelligence as a vital component in enhancing customer analytics, fraud detection, risk management and other activities.
But executives seem hesitant to adopt AI for customer-facing virtual assistants despite the popularity of smart speakers like Amazon’s Alexa or bots such as Bank of America’s erica.
A recent technology-related survey of 97 bank executives by American Banker found that only 5% have extensively deployed AI-powered virtual assistants.
Observers partly attribute this to the complexity involved with creating a meaningful virtual assistant experience for customers. That might become less of an issue as more banks transition to cloud-based platforms, but the limitations of legacy systems might cause executives to shy away from the feature.
“For small banks, it’s an enormous journey to think about how to do this,” said Greg Simpson, Synchrony Financial’s chief technology officer and AI leader. “You can’t just go out and buy a chatbot and it works. You need the infrastructure to make the kind of connections needed to make it all work. It can very difficult to do that well.”
However, Synchrony and Wells Fargo each have begun to experiment with incorporating AI into virtual assistants, and their efforts offer a road map of sorts for others willing to give it a try.
Aiming too low?
Before jumping into virtual assistants, institutions should define conversational banking in 2019.
Sherry Comes, a managing director and a leader on Deloitte’s AI & analytics group, said the industry does itself a disservice when it minimizes the virtual assistant to just a chatbot.
“A chatbot is trivial to what we’re doing,” said Comes, who spent 16 years working on IBM Watson. “You’re creating an intelligent assistant using data-driven decisions with AI and machine-learning models.
“I think the terminology is what’s tripping people up,” Comes said. "When I think of a chatbot, I think of a high school or college student in their room creating a program that just answers questions. That’s absolutely not where we are in this space.”
The industry is at a point where banks have multiple options for how and where they deploy conversational banking experiences, which adds to the complexity of the decision-making process.
Facebook Messenger for the past three years has provided banks with an entry point to AI-enabled chatbots via an application programming interface. The same is true for Alexa, which can be found now in everything from car dashboards to refrigerators.
While those third-party platforms have limitations, Comes said they nonetheless provide an option consumers might already be comfortable with.
“The last thing a customer wants is another app on their phone that they have to interface with,” Comes said. “Banks know that, so they’re embedding conversational-banking capabilities” in devices customers already use.
However, a handful of banks have shown that an in-house approach to conversational banking may be the best way forward.
A true assistant
Chuck Monroe, the head of artificial intelligence enterprise solutions at Wells Fargo, envisions a future in which virtual assistants act as a lifeline to customers for simple transactions, as well as more complex tasks. The bank is piloting a virtual assistant within its mobile banking app that aims to be as flexible as possible.
“Our strategy is about creating a customer interaction that will include a chatbot component, but will also include a live agent component,” said Monroe. “It’s about creating that seamless interaction and letting the customer decide how they want to be serviced.”
How that plays out in real-time will vary based on the situation.“If the customer has a question, maybe the bot answers it first, and then the customer service rep picks it right up” to continue the conversation, said Monroe. “We’ll have to figure out notifications to let the customer know the difference between a bot and a person,” he said.
Monroe said conversational banking must strike a balance between automation and the human touch.
He referenced a recent television commercial where a company promises human interaction when consumers contact the company about customer service issues.
“Is that a reflection that the chatbot hasn’t lived up to expectations?” Monroe asked.
He said there are some scenarios when a virtual assistant might not be the ideal option. Wells has an internal virtual assistant for employees who work within the estate care group to help customers deal with the finances of a deceased family member.
Monroe said the idea for this virtual assistant is that it enables a customer to deal with one representative throughout the duration of a single phone call.
The customer service representative can ask the chatbot questions about a credit card, home loan or other matter to enable the estate executor to stay focused on one person.
“It’s about creating those virtual interactions that are not highly emotional interactions because those are the ones where we want that human contact,” Monroe said.
And deciding how and where to draw that line may be what is giving many banks pause.
“The evolution of what the right use cases can be is causing organizations to pause and be thoughtful about creating the right customer experience,” said Monroe.
Synchrony has demonstrated that conversational banking interactions can benefit companies and their customers.
The company, based in Stamford, Conn., is in a unique position because it partners with more than 20 retailers that use its AI-powered assistant, Sydney.
The digital assistant has handled millions of chat interactions over the past year. Some 450,000 chat interactions happen per month, and they occur around the clock, Synchrony’s Simpson said.
In turn, that helps live customer service agents deal with “high-value solutions”, according to the company.
Simpson said Sydney relies on years of Synchrony call-center data to help cardholders find answers about account openings, payment options, login issues and rewards programs.
Like Wells, Synchrony has relied heavily on cloud computing to power Sydney.
“We invested in a cloud strategy and an API strategy to enable a lot of data in our legacy systems and make that accessible to a chatbot,” said Simpson.
Synchrony created Sydney as a direct response to how cardholders wanted to interact with its retailer partners, Simpson said.
“Consumer preferences have changed as far as reaching out for customer service,” he said. “Dialing into a call center used to be at the top, while email and text were at the bottom. That has flipped.”
Banks can take advantage of this shift, but that will depend on whether conversational banking fits well into an institution’s future plans.
“There is so much opportunity that organizations have to pause and say, where are we going to start with AI that has the most impact within our organization,” Monroe said.