History of Neon — Our Customer Service Chatbot
How we are building, scaling and continuously improving our Customer Service chatbot at N26.
Conversational interfaces allow users and computers to interact with each other following the principles and rules of human-to-human conversations. Thanks to the recent advances in Artificial Intelligence and Natural Language Processing, these interfaces are now getting more and more user-friendly, accurate and efficient.
At N26, we saw an opportunity to use these technologies to create and launch Neon, our own Customer Service chatbot.
Improving banking experience with AI
Our primary focus is to make our users’ lives easier by providing a smoother banking experience. This can be achieved through a simple & transparent offering, a great product, and easy access to information & services.
When it comes to the latter, we initially decided to focus on the “usual” channels — like our Support Center, Web Chat, Phone or Email, to serve customers.
Then, as part of our continuous focus on customer experience and innovation, we decided to explore the opportunities offered by conversational interfaces and chatbots, as a new way to provide support and answer the most common customer questions on any platform (web or mobile app), any day at any time, in all the languages of our different markets.
Meet Neon, 2 years old, speaks 5 languages and knows all about banking
After several months of preparation and fine-tuning, Neon was officially born in 2018. 🎉 Since then, we have been constantly progressing in the chatbot’s capacity to understand and answer our customer’s questions. And at the age of 2, Neon is already able to:
- Speak 5 languages — English, French, German, Spanish and Italian
- Understand more than 45 different customer questions about our products and services
- Handle thousands of customer enquiries every day (and every night)
- Automatically solve one fifth of them and get an explicit positive feedback from our customers for one tenth of them
They also become smarter and smarter when dealing with these requests by:
- Conducting dynamic conversations in natural language
- Understanding customers’ profiles — to provide more specific and contextualized answers
- Transferring the conversation to a human colleague when they require external help
And this is just the beginning, as Neon is continuously improving and learning new knowledge & skills!
Growing in the best development conditions
Raising a gifted child like Neon requires a very specific environment, in terms of skills, environment and Education principles.
First of all, Neon is the result of a co-parenting effort, through a multidisciplinary team that mixes expertise from Product and Technology (driving global roadmap, integration with our platforms and corresponding UX & UI), Data Science (to build and improve the required AI algorithms and backend), and Operations & Content (to ensure scaling and continuous improvement and nurturing of Neon’s capabilities).
They also rely on a state-of-the-art technology stack:
- The latest AI algorithms, enabling to understand customers’ questions (Natural Language Understanding) and to manage conversations, based on Rasa’s open source framework and relying on deep learning (e.g. LSTM and Keras)
- A deep integration into our frontend and backend environments, to provide a seamless experience when getting in touch and discussing with Neon, or when getting transferred to a human colleague
Finally, they require specific principles of education based on:
- Constant focus on value — both in terms of efficiency and customer experience
- Incremental growth — learning from past chatbot overhype by starting small and improving step by step
- Continuous monitoring of Neon’s results — from a Business, Machine Learning and Tech perspective