29-08-2024 10:43

Our customers have doubled their use of our chatbot in two years

Nova, our chatbot, is expected to engage in more than seven million customer chats this year. According to Anders Nicander, Head of One Digital, the quality of chatbots has improved significantly within the last couple of years.

Nova Nordea's chatbot — A question to Nova is typically raised in 20 different ways,

Since 2017 our chatbot Nova (Nordea Virtual Agent) has been doing its job, supporting customers with quick answers to their banking-related questions. With close to seven million chat sessions expected this year alone, Nova is indeed popular showing a 100% increase in 2 years. Anders Nicander, Head of One Digital explains:

“This is good news as we want to make banking easy and accessible, and provide our customers with the freedom to manage their finances and get answers to their questions when it is most convenient for them. Nova can answer simple questions in two seconds, there is no waiting on the phone, and our advisors can focus on handling more complex matters.”

“Nova leads you all the way”

According to Anders Nicander, there are a number of reasons for the increase in chatbot sessions:

“The quality of chatbots has improved significantly within the last couple of years due to AI and intensive internal training. Furthermore, we introduced what we call deep links that lead you all the way. If you for example have forgotten your pin code, Nova doesn't provide you with a lengthy text instruction of where to find it. It shows you all your cards asking which card you need the PIN code for. You press the card and it takes you to the see my PIN code feature in the mobile, so Nova leads you all the way.”

Digital first: From early adaptors to vast majority

In 2017 chatbots were primarily used by early tech adaptors, young people and tech savvy people who liked to play around with the functionality. In total, it reached no more than 2,5 percent of the population. At the time, there were about 40 million logins to Nordea’s mobile bank. Fast forward to 2023 and the number has exploded to more than 1,4 billion mobile bank logins.

Norwegians seem to be having the biggest appetite for Nova with the highest chatbot versus phone ratio, whereas Danes are the most conservative in the Nordics when it comes to the use of our chatbot.

But even though Nova is already a success, the intensive development of Nova continues.

“We have highly skilled colleagues dedicated to Nova in each market, making sure that the chatbot is not only providing the correct answers. A question can typically be raised in 20 different ways, and we want Nova to make sure she understands and replies accordingly,” Anders explains.

Meet chatbot trainer Lasse Eiesland Strøm, Norway

Team Manager for a customer service team focusing on written communication with customers, AI Trainer for for the last 6,5 years:

How do we make sure that Nova provides the right answers also to tricky questions?

“To make Nova provide the right answers, follow the context of a conversation and steer it in the right direction, we need highly skilled AI Trainers with a ton of experience from our Customer Service department. They know the customer journey and pain points, and know what question the customer will ask next, even before the customer does."

"We also listen to customers’ feedback. Every year, we do a deep dive into conversations with Nova and improve accordingly. That means we continuously see new ways of phrasing a question, or even altogether new questions Nova should be able to answer.”

How do you train Nova?

“For each intent, an AI Trainer need to add at least 20 different ways that we could imagine the customer would ask a specific question. To make sure this is done as precise as possible, another AI Trainer will add at least 10 quirky/tricky ways customers could ask that same question to make the work of AI Trainer #1 foolproof."

"Based on the repetition of certain words, the connection and placement of words in a group and the manual process of adding various words and synonyms, Nova should be able to separate questions from one another. This requires continuous finetuning, though. For instance, “Why am I not able to log in?” requires a completely different answer than “I have a problem when logging in”, although both questions could mean the same from the customers’ point of view.”

Digital banking

Banking innovation

Insights