Swedbank’s virtual assistant, Nina, now conducts an average of 55,000 conversations a month and can handle more than 350 different customer questions. We speak with Lotta Lovén, head of digital banking at Swedbank, about the transformative powers of this technology.
In 2015, the Swedish retail bank Swedbank went live with its ‘virtual assistant’, Nina. Designed as a first port of call for customers visiting the website – a way of resolving their queries without the need for a phone call – the chatbot made an immediate and far-reaching impact.
Within three months of deployment, Nina was handling more than 30,000 conversations a month, with a first-contact resolution rate of 78%. Although some questions still required further action, 55% of conversations were concluded there and then, with no need for the customer to escalate their query.
Some two and a half years later, Nina is a familiar fixture on the website. According to Lotta Lovén, head of digital banking at Swedbank, the chatbot now deals with around 55,000 questions a month, of which 85% are answered on the first try.
“We have a few people continuously working on increasing the number of questions it can answer, by looking at what we see most frequently, so it becomes smarter and smarter the more we use it,” she explains. “Right now it does the work of approximately 30 people in the contact centre.”
Although Nina is fully automated, customers are unlikely to feel as though they’re talking to a robot. Because it simulates a human conversational style, the experience is highly natural and intuitive; a testament to how quickly artificial intelligence is evolving in the customer service space.
“Technologies like this will definitely grow in future, and not just in the banking industry – they will grow in general,” says Lovén. “We already have examples of what the next steps might look like, in terms of Amazon Alexa for instance. But in banking, this is a strong tool for building relationships with our customers.”
The digital answer
The backstory to Nina is one that will strike a chord with many banks. In 2014, Swedbank dealt with 3.6 million inbound interactions, of which more than two million were transactional queries (i.e., the sort of simple questions a chatbot might easily resolve). Its 700 contact centre agents, employed across five sites, also generated one million outbound responses.
At the same time, the bank was seeing an upsurge in digital engagement, with 58% of its customers banking digitally and 80% logging into a digital channel at least once a month. The bank projected that by 2018, online and mobile channels would be customers’ primary mode of contact.
Swedbank felt its contact centre agents were spending too long handling basic inquiries, which in turn meant less time spent on value-added tasks (i.e., selling products). The solution seemed obvious: a self-service option that could answer customers’ queries across its digital channels.
“We were getting a lot of recurring questions via telephone, like how much do I have in my account, where can I find the IBAN number,” explains Lovén. “We saw that by introducing a chatbot, we could help our customers to get a very fast response while decreasing the number of simple questions to tellers.”
While automated tools of this kind are often viewed as a replacement for human labour, with all the attendant fears about job losses, Swedbank didn’t want to reduce any manpower. It simply wanted to maximise its resources, giving its contact centre agents more time to, say, sell insurance, as opposed to helping customers download their mobile apps.
The bank decided to team up with Nuance Communications, the speech technology company behind Siri. Described by Forbes as ‘arguably the most advanced speech recognition company in the world’, Nuance has developed solutions for nearly two thirds of Fortune 100 companies and maintains one of the largest libraries of speech data in the world.
The company has also developed many web-based solutions, which seemed the most applicable to Swedbank’s needs. While the bank had originally intended simply to build a knowledge database, Nuance proposed a different solution, Nina.
This tool is fully integrated with the bank’s contact centres and answers many questions that would previously have required human input. It can also be used internally by the customer service agents, to help them with information searches during a call.
“I would say you can compare Nina to Google – instead of having to search our website or waiting in a long queue for the telephone bank, you can get forwarded to an answer directly,” says Lovén. “So it has sped up the service and made sure the customers very quickly and conveniently get answers to the most frequently asked questions. We have always been rather forward-looking in terms of how to improve our customer interactions with the latest technology, and we saw we had a great opportunity here to support our customer support centres in their daily work.”
One of Nina’s main advantages is its ease of use. When the customer visits the Swedbank homepage, they will see, in bold lettering, the text ‘Hi, how can we help you?’. Beneath that, they can type their question freeform in a box reading ‘Feel free to ask your question here!’. Nina will then respond with an answer (e.g., a few short paragraphs or a link to a relevant webpage), or a further question to clarify what’s been asked.
Since the tool was first piloted in October 2014, its capabilities have steadily expanded. It is now capable of handling over 350 questions, up from 100 in the early stages, and the list continues to evolve in accordance with what customers are asking. What’s more, because it uses a natural language processing (NLP) technology, the questions can be phrased in any way the customer likes.
Swedbank’s Martin Kedback has described the underlying database as a “living thing”, adding that, were he to roll out Nina again, he would start with a smaller customer base first, to give the chatbot more time to learn and grow the database.
Rise of the chatbots
Evidently, this is a world apart from the stiff, formulaic interactions associated with previous virtual assistants. If we think back to Clippy, the much-reviled Microsoft Word paperclip, not only was his ‘help’ based on a very limited script, but he failed to register user input or adapt.
Nina is more user-friendly – not least because it remains a work in progress. As well as growing the database, Swedbank plans to expand Nina beyond the website and into its customers’ internet banking and mobile apps. According to Lovén, the bank has “a very ambitious agenda”.
“We want to make sure Nina’s implemented in our mobile app and our internet bank to be where it’s most relevant for the customer,” she says. “Of course there is an element of technology development here, but it’s more about utilising the possibilities we already have. We are also talking about having it voice controlled and connected to robotics, but that is not the big thing – the big thing here is to increase usage and implementation.”
For instance, customers may soon be able to log into their mobile app, and tell Nina they want to transfer 1000 Krona between accounts. Or they may realise they’re struggling with a certain functionality, and ask Nina how it can be resolved.
“We are also starting to evolve the usage into internal support functions, because we have a lot of different functions where this kind of tool could give us a much faster response and decrease the time spent in queues,” Lovén adds.
Swedbank, of course, is not the only bank to be looking into services of this nature. In Sweden especially, where branch closures continue apace and customer satisfaction scores are low, there has been a real push towards virtual assistants. Nordea Bank has the Nova chatbot in its life and pensions unit, while SEB has two chatbots named Aida and Amelia. A recent Bloomberg headline proclaimed, ‘Your banker is always in: Sweden rolls out the robots’.
This trend will most likely continue, both in Sweden and elsewhere. As machine learning technologies become more sophisticated, and chatbots harder to distinguish from humans, many banks will come to view these tools as a key prong of their customer service strategy.
That said, any mention of automated service is likely to provoke discomfort in certain circles. Can chatbots truly replace a human interaction, in a way that users don’t begrudge?
Lovén feels that, far from hurting banks’ relationship with customers, tools like Nina actually stand to improve customer satisfaction and personalisation.
“I think we need to look at customer behaviour – today we only meet about 10% of our customers physically each year, with the rest preferring digital services,” she says. “Digital services from banks have mainly been based on transactions and process-related topics, but we now have the possibility of adding on a layer of communication and interaction, meaning the digital service becomes personal.”
She says that customer feedback to date has been positive, adding that chatbots don’t take anything away from the customer – they simply add another way of communicating with their bank.
“If you don’t get an answer to your question, of course you can speak to a real person after that,” she points out.
This article appears in the Winter 2017 edition of Future Banking