The Customer Effect
Swedish bank SEB is using a ‘cognitive agent’ for customer service
- Sweden's SEB bank is part of a group of large banks and insurers using cognitive agents, or virtual humans, to handle customer interactions.
- Analysts say cognitive assistants, despite being more intelligent than chatbots, will still need to refer some inquiries to humans.
While a customer service experience of the future may involve automation of some kind, the help of a virtual human called a cognitive agent may be the next evolutionary step. Sweden's SEB bank is one of a handful of large banks and insurers (including UBS and insurer VGZ) that have enlisted the help of cognitive agents to deal with customers. The bank is working with software company IPsoft to get a cognitive agent called Amelia to act as a front-end customer service agent. Cognitive agents are virtual assistants that can supposedly think and act like humans, handle complex interactions and learn from situations. "The business case about efficiency is not the main interest," said Nicholas Moch, head of innovation, strategy and architecture at SEB bank, speaking at the Digital Workforce Summit in New York Thursday. "It's about scalability and the fact that we have customer service [solutions] that we can roll out quickly." The bank said its use of a cognitive agent began as a pilot project for company's internal help desk in September last year. By December, SEB expanded it to include front-end customer service interactions including account openings, cards, booking meetings and delivering branch office information. The bank's digital employee, who is named Aida, has an interface resembling a chatbot, but the bank claims she is more powerful. Moch said he expected employees and unions to be nervous about Aida's capabilities to put human customer service agents out of work, but found that the ability to outsource inquiries to Aida let them focus on complex interactions with customers. He stressed that Aida's objective is not to replace humans, but to deal with their questions more effectively, referring them to humans when they are needed. "Escalation [to humans] is something she should do -- 30 percent of the time she may escalate to humans and that's something we've asked her to do," he said. Customer service experiences in banking are often synonymous with an automated voice, which can be a frustrating experience for a customer who is compelled to reach out to a human to resolve issues. Chatbots were thought to solve those problems, but as Tearsheet reported in March, industry players -- including Capital One and Mastercard -- acknowledge the limitations of chatbots, particularly their ability to deal with more complicated customer inquiries. Chetan Dube, IPsoft president and CEO, said Amelia can go a lot farther than a chatbot, with its ability to mimic human thinking processes. Chatbot technology, he added, is far from this point. "Chatbots and IVRs [interactive voice response] have failed," he said. "No matter what lipstick you put on that IVR pig, it will only be a maze that will give you a sub-par experience." It's up to the banks to determine how customers interact with Amelia, he said. While SEB uses a chatbot-type interface, others have displayed a live digital human with a voice, and the technology can be accessed from a multitude of channels, including desktop and mobile. When asked if Amelia is at risk of algorithmic bias when making decisions, Dube said that the use of reinforcement learning helps her learn from situations and pivot when necessary. "There's two types of learning, that which is observed from rules, and there is also event-based episodic learning -- if you have episodic learning, agents can learn very adeptly." To analysts, the debate around whether cognitive agents like Amelia will displace humans is misplaced, as the future will be about finding ways for humans and cognitive agents to work together. "The real opportunity is in the middle when humans and machines come together -- that's where jobs are being created," said Paul Daugherty, chief technology and innovation officer at Accenture. Photo: A cognitive agent developed by software company IPsoft, is said to think like a human to resolve service inquiries.