Chatbots have been a hot topic in banking for a while now. Even though the technology has moved beyond the highest hype point, the jury is still out on whether employing this software makes sense.
It may seem ironic to some, but for many digital-first banks, live-support chat is arguably the most important channel. InSync — a mobile bank inside Alfa-Bank Belarus — interacts with 20% of its users through chat thanks to a ten-person support staff, available 24/7 to talk with customers. At the end of the day, it is these people who shape the image of the bank.
Launched in 2016, InSync has a simple goal of being close to the customer and talking in their language. It’s been doing a good job at that so far: Over the past three years, the number of mobile bank customers has grown fivefold to about 200,000 users. What’s even more important, though, is that customers of the mobile bank account for 60% of the deposits and 15% of the loans of Alfa-Bank Belarus every month.
“More than 30,000 customer s are handled through live chat every month.”
The chat support staff — affectionately known as “Alfachs” — now handle both InSync and other Alfa-Bank customers, though the former account for some two thirds of the s. The most popular questions coming through chat usually belong to five topics: functionality and usage tips related to InSync (more than half of questions); payment cards; current accounts; loans; and utility payments.
More than 30,000 customer s are handled through live chat every month, compared to the industry average of about 600 (according to Zendesk). Overall, about a third of the bank’s total of 75,000 support requests come through chat.
3 Reasons Why It’s Premature to Roll Out a Chatbot
The pros of adding chatbots to the customer service mix would seem obvious: the technology is said to be able to cut up to 85% in salary costs and handle many more customers at the same time than even the most talented customer care representative. Moreover, previous stories on 36kr have illustrated convincingly that quite a few people don’t mind talking to chatbots at all. In fact, some even prefer that out of privacy considerations.
Nevertheless, not a single chatbot is used by InSync — at least, not yet. Why would a digital-first financial institution avoid using such technology? There are three considerations of why the bank thinks it’s too early to employ this technology in banking.
1. Bots Lag in Decision Making
Before chatbots can become a viable solution for banking, their decision-making power must be improved. What good are customer support representatives — no matter a bot or a human — if they don’t have enough decision-making power to solve the customer’s problem? You can make as many good jokes as you want in a chat, but if you can’t, say, refund a fee or reverse an erroneous transaction, all you’ll have at the end is an unsatisfied client.
Trying to solve this problem, InSync has given its “Alfachs” broad freedom in terms of the decisions they can make on the spot. This approach works long-term: the bank has significantly increased customers’ loyalty and satisfaction. But how would this work with an artificial intelligence (AI)-powered chatbot? Would it be able to make decisions as well as a human operator? Would it be easy for consumers to game it in some way? Those are the questions the industry has yet to find the answers for.
2. For Now, Humans Are Better Differentiators
One of the more realistic theories about the future of banking is that at some point all banks will have pretty much the same customer offerings and same channels of communication with clients. Being available on Facebook or Twitter could have been a differentiation point a few years ago, but not now. That’s the way client communications tend to evolve in the digital age.
With that in mind, it’s timely to ask: In what way could your bank stand out of the crowd for consumers in the times of homogenization of the industry?
For InSync, the answer is the language it speaks with the users — their own language both figuratively and literally.
“At the moment, no one can understand a human better than another human.”
Even though we are all firm believers in technology in general and AI in particular, the InSync team maintains that at this moment no one can understand a human better than another human. That’s why the mobile bank employs humans for its chat function and lets them be themselves while talking to customers.
3. For Some, Language Can Be a Barrier
The third point, which is relevant for InSync but less so for many other banks, is that conversational AI in languages other than English and Chinese is simply not up to the task. Most of our bank’s clients are Russian-speaking, and the team has come to the realization that deploying a bot for the support chat would essentially mean training it at the expense of the bank’s own customers.
Setting up a chatbot requires quite some preparation and “pre-learning.” For each banking product, we would need to create a list of questions we would expect to be asked along with a knowledge base answering all these questions. Then we’d teach the AI behind the chatbot to find proper answers based on keywords and key phrases. Afterwards, however, fine-tuning of the AI model will always require interaction with customers. The more people interact with AI, the better answers it will provide.
Getting to an acceptable level could take quite a while, and InSync has no inclination to test the patience of the clients who need help here and now.
The Solution Is In The Middle
All in all, a hybrid model looks the most usable and future-proof of all possible implementations of chatbots in banking. The idea behind it is that the AI would only handle the conversation with a customer up to a certain point and seamlessly hand it over to a human support rep when it can’t be sure of the right answer anymore.
This approach is already used by some banks, even though the seamlessness part is yet to be fully solved. Bank of America may be close. It’s hybrid chatbot, Erica, talks regularly to more than four million customers.
Depending on market maturity, it’s safe to guess that we will see more high-quality hybrid chatbots within the next few years. The future looks quite bright for them, since the final goal can be achieved through a series of incremental advances rather than a major breakthrough that would require a great deal of resources to implement.