Swedish banks, including SEB, Swedbank and Nordea, are adopting artificial intelligence (AI) to increase their competitiveness.
Aida, Nina, Nova are three of the bank’s new employees. But instead of sitting behind a desk, they offer virtual assistance to internal and external customers.
“AI consists of using robotics and machine learning in combination with chatbots to automate a lot of the things we do, ”said Mattias Fras, head of robotics strategy and innovation at Nordea. “Digitization hits us as much as any other business right now. We are welcoming new banks that offer similar products at a much lower cost. AI can allow us to be competitive in both customer experience and efficiency.
This reflects the changes taking place in the Nordic banking market. Consumers increasingly want 24/7 service, while fintech companies offer competing digital banking products. Then there is new regulation, including the revised EU Payments Directive, PSD2. All of these factors change the use and access to bank data.
AI-based virtual assistants are one way to keep up with this new order. He is the ideal employee: they work 24/7, never get bored and always react quickly. By taking on simple customer service tasks, both internally and externally, they free up bank employees for more complex customer tasks and inquiries.
Swedish banks are accelerating these plans. Last year Nordea launched chatbot Nova for its life and retirement unit in Norway, SEB introduced virtual assistant Aida and retail bank Swedbank welcomes customers to its website with virtual assistant Nina.
But bank virtual assistants weren’t trained overnight. Swedbank started deploying Nina in 2014 and SEB’s Aida follows on from its first chatbot, Amelia (built on IPsoft’s AI software of the same name), which launched in 2016.
“We started to explore Aida [then called Amelia] at the end of 2015 internally with our IT departments, ”said Erika Lundin, head of SEB’s center of excellence for Aida. “We found that, for example, unlocking a Windows account on a computer was a question often asked at our service desk. So we started with some typical IT related questions. We did a pilot with 600 employees at the start of 2016. ”
At the start of 2017, Aida was deployed in SEB’s internal help desk and hired as an “intern” for the bank’s front-end customer service. Today, she handles about 13% of all IT support questions, while helping bank customers with card issues, account requests, and meeting reservations.
At Nordea, the starting point was robotic process automation (RPA) in 2015. Today the bank has around 150 software bots managing over 300 simple, rule-based processes. Chatbots entered the scene in early 2017 and Nova launched in June of last year.
Nova uses technology from the Norwegian startup Boost.ai and can handle approximately 2,000 different customer inquiries on Nordea’s website and mobile services. Fras said that the chatbots built by the bank internally which is currently being tested and is also working with suppliers.
“We are working on parallel tracks,” he said. “We use both cloud-based solutions and on-premise systems to be fast. We believe the cloud is the way to go, but there are still challenges for the bank to really enter the cloud, which is why we have a multi-system approach to chatbots.
Lost in translation
Although Aida, Nina, and Nova are continually developing their skills – including sensing emotions – their abilities still lag far behind those of their human colleagues. Language, in particular, turned out to be a problem. Most AI assistants are happy to converse in English, but offering similar levels of service in local Nordic languages is another matter.
“This is an area of AI that is not yet very mature,” said Fras. “That’s why we work with startups, like a Norwegian company that knows the local dialects. “
SEB himself taught Swedish at Aida. In 2016, the virtual assistant became the first non-English deployment of IPsoft’s AI software, although Lundin described his early language skills as “a kind of Swedish Google Translate”.
“We must have spent a lot of time developing the Swedish language with the provider,” she said. “Now Aida speaks Swedish fairly well and has around 250 cats a day. She can answer many different questions about our products and services.
Lundin believes the next step for AI is to provide more personalized service. For example, Aida does not currently identify the customer she is helping, but in the future, she could leverage detailed customer data to provide more personalized advice.
The cost of progress?
As these digital employees get more advanced, so does their deployment. At Accenture Banking Technology Vision 2017 investigation, 76% of those surveyed believed banks would deploy AI as their primary method of customer interaction within three years.
Deutsche Bank CEO John Cryan went further in a speech in 2017, stating that a “large number” of bank employees would lose their jobs as a result of the new technology takeover.
But Martin Kedbäck, product manager at Swedbank, said the human colleagues of virtual assistants didn’t need to worry about their jobs just yet.
“We see this technology more as a change in skill where the [virtual] assistant will manage not-so-complex customer journeys where humans add little value, ”he said. “This will give our advisors more time to talk to our clients where it really matters. “
Despite these reassuring words, change is underway. In October 2017, Nordea announced the loss of 6,000 jobs due to changes focused on digital transformation. Fras de Nordea stressed that robotics and AI were not a direct reason for the layoffs, but acknowledged that what he and his team are doing now will have consequences for the bank’s future workforce.
“Sometimes it’s easy to understand what kind of jobs will be lost because we are already seeing it in some of the more simple and routine jobs,” he said. “But we are also seeing new jobs being created. Coders, data scientists – we need to recruit more of these people. No one knows what the next step will be, but banking, like any other industry affected by digitization, will become increasingly automated. “