Since 2016, we have witnessed an increase in the adoption of chatbots, with banks being early adopters of the technology. The earliest usage of chatbots in financial services was in the early 2000s through text messaging. These chatbots were rule-based and could perform simple tasks like showing the account balance when instructed with a specific command. With AI becoming viable for enterprises, banks started turning to smarter automation to meet the growing expectations of customers. According to a report by Gartner, it is estimated that 85% of enterprises will perform customer engagement and relationship management with the help of AI chatbots by 2020. Banks are leveraging conversational AI interfaces to raise the bar of customer experience. These platforms also come with advanced Machine Learning (ML) and Natural Language Processing (NLP) capabilities, enabling banks to mine vast amounts of data and provide highly targeted and personalized offers to customers.
In a report that studied the impact of artificial intelligence (AI) on retail, Juniper Research estimates that retail sales arising from chatbot interactions are set to reach $112 billion by 2023. The firm also noted that chatbot interactions are set to skyrocket from a forecasted 2.6 billion in 2019 to 22 billion in number by 2023. Understandably, the outlook for chatbots in financial services looks promising, but there have been instances of successful high-profile integrations and a significant number of failed projects. Although companies are embracing chatbots, some of the AI-powered platforms are failing, and early adopters have dropped them due to disappointing results. Many chatbots fail because the companies do not clearly define their purpose and the use cases that they are deployed in are fairly broad and generic. Rather than generating a fleeting brand and PR value, banks need to clearly define their chatbot’s purpose and communicate it to users.
In one of our studies undertaken earlier, MEDICI had deep-dived into the chatbot initiatives of major banks. We have updated the below graph to reflect some of the new additions to the list. A total of 36 of the 56 banks and FIs that we researched have deployed a native chatbot on their mobile/online channels; 18 banks and FIs have deployed their chatbots on Facebook messenger (e.g., American Express, ING, RBC, etc.).
In the next section of this article, we will delve into some successful deployments.
Bank of America – Erica: Bank of America’ Erica chatbot debuted at the Money20/20 conference in October 2016. The chatbot was launched not only to serve as a virtual assistant but also serve as a customer’s “personal advocate” by sending notifications to customers, providing balance information, sharing money-saving tips, providing credit report updates, facilitating bill payments, and helping customers with simple transactions. Communication with Erica can be done by voice, text, or tap and gesture. Since its launch, Bank of America has added further capabilities to Erica, including proactive and personalized insights. As of early 2019, Erica has surpassed 6.3 million users and has serviced over 39 million customer service requests. These numbers are a significant jump from 4.8 million users and 32 million interactions in Q4 2018. In its Q1 2019 earnings, BofA highlighted that investments in its digital platform have been paying off in the form of increased engagement. From the 37 million active digital banking users, 27.1 million of whom use mobile banking – this is a 9% annual increase in active mobile banking users, from Q4 2018. The company also saw healthy engagement on its digital channels, with 1.5 billion mobile logins, up from 1.4 billion mobile logins in Q1 2018. Most importantly, 51% of all digital sales came from mobile, up from 26% in Q1 2018.
HDFC Bank – EVA: India’s largest bank by market capitalization, HDFC Bank launched EVA, the first AI-powered banking chatbot, in 2017. Eva was built to provide customers with information on HDFC’s products and services instantaneously. Within the first few days of its launch, Eva had answered over 1 lakh queries from thousands of customers from 17 countries. Eva can assimilate knowledge from thousands of sources and provide answers in simple language in less than 0.4 seconds. Eva has also integrated with Google Assistant and Amazon Alexa. Since its launch, Eva has answered more than 5 million queries from around a million customers, with more than 85% accuracy. Eva holds more than 20,000 conversations daily with customers from all over the world. The impact of the bank’s digital strategy has shown an uptick in overall digital accounts transaction, which is 85% of the bank’s total transactions. Offline branch transactions are down to 8% with ATM transactions at 6%. Out of the 85%, mobile accounts for more than 45% and the website stands at 40%.
Swedbank – Nina: As one of Sweden’s largest retail banks, Swedbank’s foremost challenge in 2014 was to enable 3.6 million customer interactions per year (2 million were basic transactional queries) to be quickly resolved through self-service on digital channels. These million inbound interactions were being handled by the bank’s 700 contact center agents (spread across five sites), which included having to deal with 500,000 emails and 10,000 social media interactions, and generate one million outbound responses. To help address this challenge, Swedbank launched its virtual AI assistant ‘Nina’ in collaboration with Nuance. Within just three months of being deployed, Nina was averaging over 30,000 conversations per month with a first-contact resolution of 78%. This resulted in two million out of Swedbank’s 3.6 million annual calls being resolved through easier self-service, freeing up time for the bank’s 700 contact center staff, so that they can focus on other value-added services.
digibank by DBS: Headquartered in Singapore, DBS Bank is a leading financial services group in Asia, with 280 branches across 18 markets. In a bold digital transformation move, DBS Bank launched an entirely new mobile-only bank ‘digibank,’ which was underpinned by an AI virtual assistant. The AI platform has extended the customer experience across geographies, channels (including mobile app, website, and Facebook Messenger), and different languages. The chatbot uses a combination of supervised and unsupervised learning strategies to train its AI models. With this adoption, DBS was able to handle 80% of its inbound inquiries by its virtual assistant and fast-tracked a new account opening in just 90 seconds – this also resulted in 1.8 million new customers for digibank India.
While there have been some ‘feel-good’ case studies on the impact that chatbots had on banks and their customer relationship management, there are numerous examples of chatbots being shut down. One of the high profile shutdowns in 2016, was when Facebook decided to shut down its automated text-based virtual assistant M, which was based on its Messenger service. According to estimates, M never surpassed 30% automation. Tencent had to pull down its chatbot because of the bots criticizing the ruling Communist party. The bots were available on a messaging app which had more than 800 million users before apparently going rogue. In 2018, Norwegian digital bank Nordnet – which has about 765,200 active customers in the Nordic region and a total savings capital of SEK 286 billion – decided to part ways with Amelia, its virtual assistant, after a year due to underwhelming performance.
As banks implement AI projects in different directions, it is imperative to separate the marketing story from what the implementation story is – there’s often quite a big difference between the two. While some are already seeing the impact in terms of more cross-selling of products and customer communications, it is still early to see tangible results that would allow them to decide whether the bank’s chatbot deployment has made an impact.