Nordea cites the maturity of the boost.ai platform as a key factor in launching a dozen AI-powered virtual agents to improve customer experience
Customer service presents a scalable challenge for large financial institutions. As business and customer needs grow, they must inevitably turn to technological solutions to help better manage volume, while looking to continue providing consistent and on-brand experiences.
Such is the case for Nordea, the leading Nordic bank whose operations span Sweden, Denmark, Norway and Finland, with each market receiving thousands of customer service requests daily.
Founded in 1820, Nordea today serves over 9 million private customers and more than 500,000 active corporate customers across these four markets, resulting in a varied channel mix of customer service touchpoints via in-person, phone, email and both live and automated chat.
Nordea's results at a glance...
- 12 language-agnostic virtual agents across four markets
- 90%+ in-scope resolution rates across all customer-facing virtual agents
- Nova outperformed previous in-house and external chatbot solutions
In 2017, Nordea’s Norwegian operation partnered with boost.ai to build and develop a virtual agent powered by conversational AI. At the time, the bank was receiving upwards of 2 million contacts per year across three contact center locations staffed by approximately 150 FTEs.
The virtual agent - named Nova - was realized as a mechanism to nudge customers towards the chat channel as the bank identified that its customer base preferred to interact with it this way and had the advantage of being able to handle multiple inquiries simultaneously.
“At first we had been looking at several chatbot companies but we wanted a more ambitious solution,” says Egild Homme, Technical Coordinator at Nordea Norway responsible for the customer service platform and who played a key role in procuring and implementing Nova since the beginning. “What sold us on boost.ai, was when our Head of Banking saw a demo for their chatbot with another Norwegian bank. Boost’s solution was more mature than anything else we had seen and it was like, ‘Alright, I think it’s time. We want this.’ ”
Homme and his team began building the virtual agent in July 2017 and three months later Nova was live. The process included building out the model for the virtual agent based on boost.ai’s pre-built module of common banking-related topics that Nordea could layer its brand-specific content on top of. Today, Nova in Norway can answer questions on more than 2,300 banking topics related to private and business banking (as well as the bank’s Life and Pension division), and averages around 50,000 conversations each month with private banking customers, approximately 50% of which Homme says are handled solely by the virtual agent, with no assistance from a human agent.
"For banks, AI is going to be transformative across a wide range of applications. At Nordea, we acknowledge the importance of a scalable chatbot strategy that the boost.ai platform enables. It is a key component towards human-centered digital transformation."
- Mattias Fras, Group Head of AI Strategy & Acceleration, Nordea
Nova’s primary scope, until recently, was answering general questions on the bank’s ‘open pages’, a core functionality that helped Nordea to decrease customer wait times and provide 24/7 access to information without the need to staff up. Since 2021, the bot is now more fully integrated, allowing customers that are logged into the net bank via desktop or its app to be guided to a number of self-service functions, such as blocking a card or finding their IBAN number.
“Nova has allowed us to enable better self-service for our customers,” says Olga Pavlova, Senior Business Developer at Nordea Finland, who manages the development team which provides integration between boost.ai and Nordea’s various front-ends. “For example, if a customer has forgotten their PIN code, the chatbot will provide a list of cards available for this specific customer. Clicking on one of the cards will redirect the customer to the relevant place in our mobile app where they can find or reset the PIN.”
This concierge-like experience means that customers can be guided to the information they need without having to search Nordea’s website or speak to a human. Customers have agency to quickly take care of certain tasks themselves which, in turn, frees up the bank’s support staff to focus on more complex customer inquiries that may require additional attention and care.
The road to a unified, company-wide conversational AI strategy
As of writing, Nordea currently has 12 virtual agents in production on the boost.ai platform, operating in each market’s local language (as well as English in some countries). This includes customer-facing versions of Nova in each of its key markets, as well as multiple internal support bots for functions like HR and 2nd-line support.
Boost.ai’s market-leading conversational AI was a critical component in Nordea’s decision to go all-in on a single vendor to help orchestrate the bank’s automated chat strategy. The mature nature of the Natural Language Understanding (NLU) algorithms powering the boost.ai platform ensure flexibility in deploying a single solution across multiple use cases, and the possibility to meet the scalability requirements of a bank like Nordea without compromising on resolution rates.
While Nordea has netted numerous benefits from this holistic approach to its conversational AI implementation, each of its different markets took its own journey to get there.
As Nordea Norway was spinning up its first instance of Nova in 2017, its compatriots in the other Nordic countries were experimenting with different chatbot solutions, developed both externally and in-house.
Nordea has 12 virtual agents across four Nordic markets on the boost.ai platform, assisting with customer-facing and internal support inquiries
Nordea’s teams found that, when it came to training and improving the virtual agent, boost.ai’s solution was ultimately superior and more flexible than the competing chatbot platform they were using. Instead of having to engage the vendor each time they wanted to add new content, the boost.ai platform allows Nordea’s AI trainers to update content on the fly, ensuring that Nova is always ready to help customers with the latest product and services information.
“Our agreement with the other Nordic countries was that if we were ever in a better shape with a different vendor, we would switch to the platform that was the better choice,” adds Homme. “As time passed, we realized that boost.ai was outperforming the other versions of Nova, particularly when it came to the speed of training which is crucial. The more training, the smarter and better it gets, resulting in a better customer experience.”
Similarly, with the chatbot that Nordea had built in-house, it was decided that, while it was a good example of the bank’s internal innovation initiatives, it would require significant resources to continue developing in the long term. As Nordea’s automated chat requirements would only continue to increase over time, it made sense to go with a dedicated conversational AI platform rather than have Nordea’s engineers build out various advanced functionality themselves.
This led to Nordea Sweden, Denmark and Finland’s customer service organizations adopting the boost.ai platform in full and launching their own versions of Nova in March 2021 across both web and mobile platforms.
Building a virtual agent three times smarter
Part of Nova’s success has been the open line of communication between the chatbot teams across Nordea’s markets and boost.ai’s core Financial Services team. The experience garnered from developing Nova in Norway was a valuable resource to the other three countries when they began building their own versions. It offered a chance to start from scratch, while still leveraging the expertise from their past experiences as well as that of both the Norwegian and boost.ai teams.
The software behind boost.ai’s virtual agents helps fuel collaboration within Nordea’s AI trainer teams. They can review daily conversation metrics and customer feedback, as well as utilize a suite of conversation building and training tools to create dynamic interactions in almost-real time.
“Having now experienced three different chatbot solutions, I would say that boost.ai is by far the easiest to build with,” says Riikka Leinonen, Chatbot Content Lead at Nordea Finland who has extensive experience with both the bank’s in-house chatbot project and a competing solution. “I like the hierarchical view where you can see in one glance how everything is connected. This let us concentrate more on creating quality interactions compared to the previous technology so that customers get specific answers instead of going through long decision trees.”
Since switching to the boost.ai platform, the AI trainer team working on the Finnish version of Nova has more than tripled the number of topics that the virtual agent can answer questions on, allowing it to truly scale alongside the growing demands of customers. Currently, the bot has an in-scope resolution rate - conversations that are marked as ‘solved’ - of 95%, where the user has found the answers they were looking for and had an experience comparable to that of a human agent having answered the same question.
Pioneering scalable automated customer service
These results are by no means unique to Finland. Across the border in Sweden, Nordea’s bold implementation approach of placing Nova as the first point of contact for customer service via chat across all its markets is paying dividends there as well.
“In Sweden, if you want to chat with us, you will always chat with Nova first,” says Eric Lindberg, Application Owner Personal Banking at Nordea Sweden. Lindberg is in charge of ensuring that Nova is compliant with the bank’s regulations, as well as coordinating between vendor relationships, IT development and business needs. “Then, if you are a specific customer, you can escalate to one of our agents if you need further assistance. We use Nova as our first-line chat support and our customers respond to it positively. Over 50% say that they get the help they need from the chatbot.”
This chat-first approach means that Nordea Sweden can efficiently automate hundreds of inquiries each day without ever needing to escalate to a human. To date, the virtual agent has a 91% in-scope resolution rate for private banking customers and 95% for corporate customers.
This is impressive when you factor in that Nova averages over 220,000 conversations per month across the Nordics, purely for private banking customers
“Nova has allowed us to enable better self-service for our customers. If a customer has forgotten the PIN code for their card, the chatbot will provide a link to the relevant place on our website where they can find or reset the code.”
- Olga Pavlova, Senior Business Developer, Nordea
Pavlova credits the overall package of the boost.ai platform, and not just the technology, as a major factor for how the bank built each version of Nova to achieve these results. “It was easy to get onboarded thanks to very good training material, plus the software is easy to use,” she adds. “If there are any changes that need to be done in production, it’s straightforward and, even once the bot is live, we can make the changes we need.”
Conversational AI has the potential to help businesses scale customer service in a way that is efficient and cost-effective. Nordea has proven that, in order to take full advantage of the benefits of conversational AI, it’s important to partner with a vendor that can deliver a mature platform with not just the best NLU, but an overall suite of features that can meet the needs of a such a large enterprise.
Nordea found just that in boost.ai and, with 12 virtual agents deployed across four markets, they are pioneering the use of artificial intelligence for digital customer service in banking, setting a high bar for how to automate at scale while continuing to deliver consistent, on-brand customer experiences.