​Swedbank humanises customer service with artificial intelligence platform

One of Sweden's largest retail banks, Swedbank, which employs a team of around 700 contact centre agents across Europe, the US, and China takes two million transaction calls every year. However, according to Swedbank head of operational support Martin Kedback, the bank was looking to reduce the amount of transactional calls that were coming through so its agents could focus on value adding to the business.

"Out of those two million calls, transactional calls took up almost all the time of our contact and branch office workers. But we wanted them to do was spend time on more value adding, relationship building activities. For example, instead of helping people transfer money from one account to another, we want our people to help them purchase a car or insurance for their mortgage," he said.

Initially, Kedback said the bank thought the ideal solution to reduce transactional calls would be to build a knowledge database. But enterprise software provider Nuance suggested the bank was going to gain more from integrating a virtual assistant named Nina.

"Nuance came back and suggested we started taking out the transactional costs and reducing waste externally, then after we done that they suggested we look at the internal knowledge database system," he said.

According to Kedback, Nina, which was implemented a year ago, was built on a knowledge database based on two sources of information. The first one was based on the history of customer internet searches made on the bank's website, and the second was based on the details customers provided when they made calls to the contact centres, such as "I want to buy a house", instead of saying something generic as "mortgage".

"The database is a living thing; it depends on what type of questions our customers are asking. We started off with 100 -- Nuance called it -- conversations and now we have 250," he said.

"It's important to note that Nina is not a Q&A database because it starts an automated conversation with you. She would ask you clarify your question in order to be specific in the answer she gives, and that conversation is controlled from a conversation tree."

Since launching Nina, Kedback said 30,000 conversations a month now go through the artificial intelligence platform, which make up approximately 80 percent of all calls the company used to receive. He cited out of those 80 percent, 60 percent are deflected calls and 20 percent are channelled inquiries, which is when customers are redirected from Swedbank.se to contact a call centre or visit a branch.

So far Nina has only been integrated into Swedbank's Swedish website, Swedbank.se. The bank is also in the process of rolling out the platform to 58 additional websites that belong to Swedbank. Kedback said there are plans to integrate Nina into the company's mobile app, in the hope that users will be able to speak to Nina in a similar manner to how Apple iPhone users are able to speak to Siri.

"So you would have a scenario where I would say: 'Transfer $100 to my wife', and she would do that for you and that's a very likely scenario with Nina," he said.

Michael Buckley, Accenture Interactive ANZ managing director, highlighted that many enterprises are beginning to realise how a customer's experience can be enhanced when platforms begin to leverage data. He noted that traditionally, platforms such as a websites focused on housing content, and that limited how customers interact with a business -- but with the rise of artificial intelligence, it can change the interaction between businesses and their customers.

Buckley pointed to brands such as Spotify and Netflix as examples of platforms that using artificial intelligence drawn from algorithms and data to improve a user's experience by recommending similar music or movies. He further added that as a result of this, the recommendations are drawing people back into continuing the experience they are having with each of those brands.

"The way content has evolved is instead of saying: 'Hey, other people like George have bought this particular clothing. You would also like that potentially?' It's now at the stage where it actually says you would like this, or we are recommending this because your behaviour was such that it would be interesting to you."