Share this post:
KBC Group NV has the core motto “We speak your language,” but in 2015, we realized we weren’t actually speaking the language of young people. Youngsters like apps, but the apps KBC had at that time were not tailored to the user experience, nor products kids expected or wanted. They were financial apps focused on the user experience of adults.
When we set out to build a kids’ banking app, we knew that money is more than a financial transaction. It’s part of a bigger conversation around things like saving for a goal and sharing success.
We wanted to create an app that was easy to use and looked familiar to kids who are already using apps including Snapchat, Instagram, Facebook and WhatsApp.
With K’Ching, an intelligent chatbot whose name is an onomatopoeia for the sound of an old-fashioned cash register, kids won’t have to listen to a parent telling them they’re wasting money on movies and take-out. They’ll hear a neutral and non-judgmental message like, “Hey, you spent so much this month that you can’t contribute anything to your savings goals, so it’s going to take you longer before you get your PlayStation.”
The K’Ching app has a familiar timeline format, so when the youngster has saved up all the money, and the big day has arrived that he can go to the store and by the PlayStation, he can share the news in the app with his friends. The app also enables users to communicate through instant messaging.
Building a chatbot with IBM Watson
Initially, there is nothing in a chatbot. It has to be trained to answer questions. In this case, because K’Ching is an open chatbot, there is no limit on the types of questions. The KBC developer team was looking for a chatbot platform that was simple, easy to manage and scalable.
It found the capabilities it needed for K’Ching in Watson Conversation on the IBM Cloud.
The developer team had built a proof of concept version of K’Ching on another chatbot platform, but when the team saw a demo of Watson and the Watson Conversation interface, it realized it was a far more robust solution. At that time, the other platform could only support only a fraction of the intents per app as Watson Conversation. The Watson platform was simpler to use and had a business interface, which made it easy for the business team to add training samples and intents and construct dialogs, while the IT people dealt with things behind the scenes such as logging and spell check.
The team made the switch, and three months later, K’Ching was in production with Watson.
K’Ching is able to answer the majority of users’ questions and can understand emojis, slang and common abbreviations that kids use. Some kids can’t believe that it isn’t a person answering them and try to antagonize K’Ching, which the developer team takes as a compliment.
What’s next for K’Ching
Currently, K’Ching speaks only Dutch, but is expected to learn French, German and English in early 2018. Since April 2017, K’Ching has answered more than 120,000 questions from more than 24,000 users primarily aged 15 to 17 years old. The top user has asked more than 200 questions. Many conversations last more than 15 minutes.
Because of the success KBC has experienced with K’Ching, there are now multiple chatbot projects at the bank. The second chatbot deployed, Harry, is being used internally by the HR department.
Integrating the bank’s back-end systems with K’Ching is one of the first things that’s on the horizon. As in the earlier example, if a bank card was ruined in the washing machine or glass of water, rather than direct the customer to a procedure to order a new card, K’Ching could initiate the process itself.
Or rather than be told how to check a balance, a banking customer could hear their balance from K’Ching.
Another idea is to give K’Ching its own social media channel, such as Facebook Messenger, which is where the first pilot of the K’Ching chatbot resided. This way, it would be possible for users to interact with K’Ching without even installing the app.
Additionally, KBC is considering implementing technologies to help K’Ching find the answer rather than reply, “I don’t know the answer.”
In my next post, I’ll be sharing how developers created an authentic conversational interface.
Learn more about Watson on IBM Cloud.