How conversational AI is pioneering the digitization of banks in the Nordics

Today’s banks are increasingly turning towards artificial intelligence as a crucial tentpole in their digital strategy - and reaping significant rewards.

As the technology continues to mature, virtual agents are now able to offer a far wider range of services than ever before, courtesy of groundbreaking advances in conversational AI. While virtual agents have already been able to assist banking customers with menial tasks and queries for some time, the potential for them to open up even further presents many exciting new opportunities.

In 2016, Scandinavian software developer boost.ai developed Banki, a virtual agent for SpareBank 1 SR-Bank, a Norwegian bank looking to expand its digital capabilities while reducing the costs associated with its growing customer service needs. Banki was built using boost.ai’s complete conversational AI solution that combines the company’s advanced knowledge of machine learning and natural language understanding (NLU) and scalability with enterprise-grade security and privacy features.

Quick ROI

SR-Bank’s digital banking advisor, Banki, has transformed their customer interactions.

SR-Bank’s digital banking advisor, Banki, has transformed their customer interactions.

SR-Bank’s digital banking advisor, Banki, has transformed their customer interactions.


In short, conversational AI has helped the Norwegian bank hit significant milestones, including:

  • Increasing customer support capacity by 149.3% without hiring any new staff

  • Handling over 23 000 conversations per month through Banki, the equivalent of 20 FTE

  • Decreasing customer interactions in error handling by 24%

  • Successfully answering 4 in 5 questions without the need for human support

SR-Bank established that to satisfy its outline of what Banki should be able to accomplish when answering customer queries, a balance would need to be struck between efficiency and quality of service.

“In order to do that, you need smart algorithms to understand what a customer’s intent actually is,” says Ramtin Matin, Lead Technological Strategist, SpareBank 1 SR-Bank. “Working with boost.ai has expanded our horizon, both in what is capable with conversational AI, but also how far you can push the union between doing business and incorporating technology.”

“Chat-first” strategy

Since the implementation of Banki, SR-Bank has begun prioritizing a ‘chat-first’ strategy, acknowledging that its customers prefer chat as a channel over phone and email. Banki is currently answering approximately 23 000 conversations a month and can recognize more than 1 900 intents, offering a response time and level of accuracy that simply isn’t possible from its human colleagues.

The introduction of Banki has dramatically improved SR-Bank’s ability to handle a higher volume of customer inquiries, resulting in a capacity increase of 149.3%. Where one customer service employee answers approximately 10 300 chat dialogues per year, Banki is currently doing the work of nearly 20 full-time employees, greatly improving the bank’s response rate and thus negating the need to hire additional staff.

A marked ROI

A particular use case in which the virtual agent has helped SR-Bank see a marked return on investment is in the area of proactive error handling. “When you authorize yourself for our e-banking solution and encounter an error, perhaps due to a network outage, the error codes are automatically redirected to Banki,” Matin explains. “It then pops up and tells the customer that the error they have encountered means they have to either try again in a couple of minutes or contact customer service.”

This type of instantly-accessible feedback presents customers with useful information rather than a generic error message and is a tangible advantage of a virtual agent over a human.

In March 2018, Norway’s primary electronic identification system, BankID, suffered from operational problems leaving customers unable to login to their accounts. During the outage, Banki successfully fielded upwards of 4 000 conversations in a single day. An unprecedented number of interactions that, without the help of a virtual agent, would have otherwise put a tremendous strain on SR-Bank’s service center.

“We have detected that within this category, the number of customer enquiries has gone down 24-per cent,” Matin notes on how Banki’s efficiency in error-handling has far exceeded the bank’s expectations. “It makes a huge difference. No one wants to answer questions about login-errors and it’s something you can easily automate.”

Support staff can then focus on dealing with more complex customer queries -  something that Matin stresses was a big part of the decision to adopt Banki in the first place. “This is not about replacing people, this is about reshaping the organization and employees from doing repetitive tasks into more meaningful work day-to-day.”

Looking ahead

So, what does the future hold for Banki? Matin says that the roadmap for SR-Bank’s virtual banking agent is long and that he wants to see Banki “move away from being seen as just a content bot, into having it serve more utilitarian functions.” There are plans in place for Banki to be licensed as an authorized financial advisor by 2019 - something that has the potential to be truly game-changing - allowing it to handle bigger-picture concerns such as assisting customers with retirement plans and mortgages.

The advanced level of conversational AI required to achieve this goal is non-trivial and wouldn’t be possible without boost.ai’s underlying software and support. Henry Vaage Iversen, the company’s Co-Founder and Chief Commercial Officer adds, “We continue to work closely with SpareBank 1 SR-Bank to help them achieve their vision of a virtual agent that can take their customer interactions to the next level.”

Bolstering brand loyalty

SR-Bank was one of the first financial institutions in Europe to integrate core banking services with artificial intelligence, helping to put it on the map as a leader in intelligent virtual banking. With help from boost.ai, it hit another significant milestone in 2017, introducing a proactive version of Banki that allows customers to complete banking-specific transactions directly via the virtual agent, something the bank says did a lot to bolster brand loyalty among its tech-savvy customers.

In a further attempt to place conversational AI at the center of its digital customer service strategy, in October 2018, SR-Bank underwent a test phase utilizing Banki as its only first-line support. Part of this test required making human chat available on the bank’s open website - previously customers were required to login to speak to a human agent - so that if Banki was unable to assist with a query, it could hand-off to second-line human support.

The test revealed that only 22-percent of conversations with Banki were sent to a human advisor, meaning that approximately four out of five customers had their queries resolved by the VA . “This shows that Banki is capable of handling most of the general questions visitors to our site have,” Matin explains, adding “A lot of people speak about virtual agents and about having technology in-house, but surprisingly few actually mention how you realise your return on investment. This is a concrete example of how technology allows us to shift the needle and become more effective for our customers.”

The human element of AI

A byproduct of upskilling staff has been the emergence of the AI Trainer, a burgeoning new profession within the virtual agent-supported customer interaction workplace. A role pioneered by boost.ai, AI Trainers are one of the key tenets of conversational AI, responsible for ensuring that the underlying systems of a virtual agent can understand how to accurately translate customer queries into actionable outcomes.

“I think the necessary skills for a good AI Trainer are to be patient, thorough and to be able to predict what customers might ask for,” says Monica Sandve, an AI Trainer at SR-Bank responsible for training Banki. While different to traditional customer service roles, she adds that there is still a need for “humans to work with the virtual agent because it can’t do it alone, you need to train it to reach its full potential.”

This is a sentiment clearly echoed by Matin when speaking about competency changes within his organization. “Many of our customer support staff have a unique skill set because they understand the semantics of how people actually ask questions,” he adds, “If you couple that with the relatively high complexity of products you have in a bank then you have the perfect fit for an AI Trainer.”

The partnership between SpareBank 1 SR-Bank and boost.ai illustrates the enormous untapped potential for organizations within the banking sector to see significant returns on their investment when implementing conversational AI into their digital strategy.