UX learnings from ATB’s AI powered Virtual Financial Assistant chatbot
There are pros and cons to being a market leader in any technology area. We have previously shared our excitement at being at the forefront with our virtual assistant chatbot powered by AI.
For our team developing the ATB virtual assistant, applying machine learning techniques to improve natural language processing is a foundational effort. Incredible by-products of that effort have been the discoveries around how people interact with the chatbot on a human level, providing learnings that are transferable across user experience design problems we are working to solve.
One of those learnings is that the chatbot has the intrinsic ability to increase the discoverability of the solutions our customers are asking for. The chatbot is able to break down barriers by providing a safe environment for customers to ask questions that can help increase their financial well-being.
To explain how a chatbot does this, consider, have you ever felt embarrassed to ask a question about mortgage terminology? Or looked for information online on how to change your debit card limits? It is human nature to feel self-conscious when one feels unaware or out of their depth. But people are often happy to ask these kinds of questions to a chatbot, when they also trust it to understanding what they are asking for.
Another aspect to this 'discoverability problem' is that customers don’t always know exactly what they need or how to ask for it.
The deployment of natural language processing (NLP) within the ATB virtual assistant addresses this problem directly. Think of the chatbot's NLP abilities as a patient and observant partner, truly listening and responding to the intention behind words rather than the words themselves. This act of listening to true intention is core to the virtual assistant and helps people discover the truth that they are looking for, in a comfortable way. Start with listening, layer in machine learning, and you will find a recipe for creating trust and engagement with users.
The deployment of supervised machine learning creates a positive feedback loop, where the NLP capabilities of the virtual assistant are constantly improving. Users themselves are constantly teaching the chatbot; each utterance contributes to its growing abilities.
So what does it really mean when we talk about truly listening to a customer's intention and not just their words? Maybe someone gives an utterance, "Hey, I lost my credit card, what do I do?" As individuals, we read this and say, 'we should help this customer make sure their card is safe.' In its own way the chatbot does the same thing.
This is a great example of the end-to-end experience, as the chatbot can't simply hear you, but is also capable of responding and helping. Our teams work on both sides of the experience, augmenting the chatbot’s NLP capabilities as well as deepening its connections to core ATB systems. Facilitating deeper, more data-driven, and personalized responses. The first iteration of a response may be simple content. If customers validate the importance of that response with their usage and feedback, the chatbot learns, allowing for stronger mappings.
The ‘North Star’ guiding the ATB virtual assistant is our commitment to helping people discover the solutions they need, when they need them. Stay tuned to alphaBeta for more details about upcoming feature releases for the chatbot and how we’re applying our learnings in user experience design to new channels.