The contact center experience often languishes at the bottom of banks' priority lists.
In a 2023 technology trends report from the research and advisory firm Celent, only 18% of bank leaders surveyed named contact center transformation as a priority for customer engagement and delivery in 2023. Only branch network transformation ranked lower. In contrast, "improving digital channels and self-service capabilities" topped the list, at 63% of respondents.
It's a common thread when Celent surveys financial institutions every other year. The contact center is consistently ranked low as a priority for funding, said Bob Meara, principal analyst at Celent. He says that's a mistake.
"The growth in digital hasn't produced a promised reduction in call center volume," Meara said. "Banks need to wake up and understand how important the contact center is and fund it accordingly."
The Seattle-based institution, known as WaFd Bank, revamped its contact center experience in July and its chatbot in October of this year. Previously the contact center required three different on-premises legacy systems to handle inbound call routing, call recording and phone banking, said Dustin Hubbard, chief technology officer at WaFd Bank and Pike Street Labs, a fintech innovation startup and subsidiary of WaFd.
"It was the conventional 'thank you for calling, the menu options have recently changed, listen carefully,'" he said. The legacy systems were not integrated, meaning that if a client, for example, wanted to check their account balance — a query that amounts to 20% of all inbound calls — they would transition to another system. The bank estimates that it would take a customer four and a half minutes to access their account balance by phone.
Relying on interactive voice response can be inefficient, as customers get frustrated and enter whatever number they need to go straight to an agent.
"It's putting an avalanche of calls to agents that shouldn't be handling them," Meara said.
Meanwhile, the online chatbot ran on a different system, meaning when stymied customers tried to escalate the conversation to an agent, the chats frequently went unanswered if the agent forgot to log into the program.
In Meara's view, the optimal modern contact center would use artificial intelligence to identify what information the agent needs and summon it in real time; the escalation workflow would ensure that when online chats are passed to an agent, they have context around the customer query; and that when customers initially call, they can converse with the voice system in natural language instead of pressing numbers to navigate a decision tree.
WaFd's changes reflect many of these best practices.
The CEO of Seattle-based Washington Federal says Luther Burbank Corp. was an attractive acquisition target because it offers an entree into California and has a commercial lending operation that could follow WaFd's recipe for growth over the past decade.
The $20.8 billion-asset bank replaced four different systems with cloud-based contact center and chatbot software from Talkdesk, combined with several technologies from Amazon Web Services including conversational artificial intelligence from Amazon Lex. If a caller is enrolled in voice authentication, an application programming interface built by WaFd will inform Amazon Lex that it is a secure connection, and the voice bot will pick up the conversation.
Hubbard said he believes WaFd is one of the first financial institutions to string together the Talkdesk and AWS technologies in this way. He chose Lex because he finds it to be the best tool in the industry at inferring people's intents, even when they don't articulate them in precise terms.
Callers can converse with the voice bot to state what they need, or be connected to a call center representative when the bot recognizes the customer needs a human agent. If the customer is not enrolled in voice authentication — a feature WaFd will soon start marketing more heavily to its customers — the bot transfers them to an agent.
The new chatbot is built on AWS and interoperates with Talkdesk, meaning inbound chats are escalated to agents more seamlessly than before. Agents now have the ability to initiate a video call and screen sharing within the chat.
WaFd also uses its own APIs to let customers check their balances and make transfers over the phone or by chat.
At the same time, WaFd has layered in Amazon Comprehend, which detects customer sentiment of voice and text conversations in real time. Agents can see in Talkdesk why the customer was escalated to them and whether a customer is feeling "negative," "positive" or "neutral."
"It gives them more perspective on who they are dealing with," said Hubbard.
Hubbard prefers that his vendors do not integrate directly with each other, which is why he has Talkdesk and Amazon Lex "speak" to each other through APIs.
Direct integration creates vendor lock-in, he said. "That makes it technically complicated to replace underperforming vendors down the road."
The new systems have also helped WaFd gather useful data about where self-service opportunities lie. With the old phone-tree method, "You don't know why customers picked something," said Hubbard. "Watching what people are saying helps us know."
The bank has already turned one of its findings into a feature in the chatbot: when a user asks for branches in their area, the chatbot will prompt them for a ZIP code. It then uses that ZIP code to look up branches in that area or, if none exist, within a 15-mile radius.
"My hope is as we add more self-service options, the more mundane, simple tasks keep getting handled by a bot and the more complicated tasks keep going to an agent," Hubbard said.
As for those customers who call WaFd to check their balance, the progress is already palpable. The time to do so has gone down from four and a half minutes to 28 seconds.
