In late 2016, the Swedish Bank, Skandinaviska Enskilda Banken (SEB), launched Aida, an artificial intelligence cognitive agent designed to provide customer service to their clients. The project started off being tested with their internal staff and later, based on its results, it was implemented externally with the bank’s customers.
Analysts and journalists have widely reported on the success of Aida as a cognitive agent for the bank. In this article, I want to explore why more banks are looking at such AI solutions, how they can improve customer interactions, and discover some of the best practices that we can take from SEB’s implementation of it.
A render of Aida, the AI female chatbot used by BEM.
Over the last couple of years, banks have widened the channels available to their customers: from traditional branch banking and extension counters to digital solutions such as mobile and phone banking. According to Citibank’s 2018 Mobile Banking Study, consumers are migrating from the traditional banking tools: 91% of mobile banking users prefer using their app over going to a physical branch, and 68% of millennials think their smartphones will replace their physical wallets.
In line with this shift towards more user-friendly experiences, traditional Interactive Voice Response (IVR) systems have become more precise at pinpointing the user’s requests, service or claim. Despite being in use since the 1960s, IVR’s still play an important role in small in large businesses:
Although IVR’s still have a place in the market, there remain some crucial limitations that drive users to other kinds of more human-like interactions. This is due to the lack of intuition IVR’s have since they are based on a predefined set of question/answer flowcharts such as the following:
Traditional IVR flowchart example
Given the static nature of IVR’s, they are usually restricted to basic banking features like:
Taking this into account, nowadays artificial intelligence powers the possibilities of traditional IVR’s. Cognitive assistants are now able to predict the intent of the customers – cutting off unnecessary interactions – such as to correctly redirect clients to the specific area, agent or department; automatically detect fraud, capture data in order to predict possible outcomes; and overall, optimize future interactions based on the data gathered.
The Swedish bank, SEB, developed Aida as a next-generation IVR that could better understand the pain points and immediate needs of the customers. At first, in its pilot stage, it was exclusively designed to assist the company’s internal help desk, however, later on, it expanded as a front-end customer service representative that could handle different complex tasks such as opening bank accounts and products (credit cards), schedule meetings and offer additional information on the bank.
According to Nicholas Moch, head of innovation, strategy, and architecture at SEB bank, the main objective of Aida is not to replace humans, but to be more effective at the time of dealing with customer’s queries. By doing so, human representatives have more time to focus on complex situations. Despite this, human interaction will always be needed whenever the system cannot resolve an issue: “Escalation [to humans] is something she should do – 30% of the time she may escalate to humans and that’s something we’ve asked her to do”.
Click here if you’d like to chat with Aida (you’ll have to speak Swedish though…)
The impact of Aida was such that it granted the Swedish bank the Supernova Award for innovative use of AI technology, which took place at the Connected Enterprise Summit in California, back in 2017.
Prior to being implemented by SEB, Aida’s predecessor, Amelia was tested at Swedish bank Nordnet. Despite working properly, the results were not as expected. As Peter Dahlgren, former CEO at Nordnet, stated in an interview: “We have tried it towards customers, and the response is ok but not overwhelming, so we are choosing to prioritize other things within our AI focus in the short run”.
Another issue that arose were language barriers. Most AI assistants can easily speak in English, however, being able to offer these same quality standards in Nordic languages is very different. According to Erika Lundin, head of SEB’s center of excellence for Aida in an interview with Computerweekly.com: “We had to spend a lot of time developing the Swedish language with the vendor…Now Aida is quite good in Swedish and has about 250 chats a day. She can answer a lot of different questions about our products and services.”
It is worth reviewing these roadblocks that the bank experienced, because they demonstrate the evolution of the technology. While the first version of the cognitive agent did not work as executives had hoped, it did provide valuable experience and learnings. The success of Aida is based on this experience, as well as the rapid maturing of artificial intelligence technologies over the past several years.
After its launch, more banks in the region have developed these kinds of solutions and services. Apart from SEB’s Aida, Nordea launched Nova, a chatbot focused on the pensions unit in Norway that can handle around 2,000 different customer questions regarding Nordea’s website and mobile services. Swedbank integrated Nina, a virtual assistant to greet visitors on the website.
These examples demonstrate the forward-thinking nature of banks in the Nordics, and provide excellent case studies and best practices for other banks across the world looking to implement new technologies, and in particular AI-powered cognitive agents.
October 16 / 2019
October 14 / 2019
October 10 / 2019