U.S. Bank Virtual Assistant
The Client
U.S. Bank
The Problem
The existing chatbot was not conversational and failed to answer many of our users’ questions.
The Solution
Create a proactive personal virtual assistant that knows you (in relation to the bank) and understands what you mean rather than what you’ve said.
My Role
Conversational Interaction Designer
What I Did
I collaborated with a cross-functional team of product managers, user researchers, business analysts, visual designers and an engineer team to develop and test early virtual assistant concepts .
About Our Chatbot
U.S. Bank had a chatbot in production for a couple years. However, it was a command-based interface. It responded to users based on a select set of questions and answers. The team wanted to develop a virtual assistant to serve the unmet needs of users with a conversational assistant in its app.
“Can I deposit a check online? I am in Florida and there are no U.S. Banks.”
“I want to deposit a check to my checking account using mobile deposit but the app defaults to a savings account. How do I select the checking account to receive?”
Scope and Script Development
The product team narrowed down a couple of use cases to test our hypothesis. From that, we develop a couple of scripts for the VA and the order of operation and talking through the steps prior to sketching our wireframes for each step of the VA expereince.
In the first sprint the product team wanted to test a standard scenario with voice; how to pay a bill. We want to learn if voice made current long and complex task faster and easier for users. We will also test the virtual assistant voices and determine which gender our participants preferred.
Design and Prototyping
We wireframed task flows to share each concept with the stakeholders and engineers for technical feedback and to understand our constraints with each solution.
The team begin drafting the steps in this process and wrote the dialog with a content strategist to match our brand’s voice and tone. I reviewed the scripted with the product manager and content strategist before sketching wireframes. Once, the wireframes were agreed upon we passed the wireframes to the visual designer to develop UI components. After that, we proceed to hand the visual design off to our UI developer to build high fidelity prototype with the script in hand for usability testing at the end of the sprint. The team worked with our research team to define the scope of our research and begin recruiting a board demographic of existing voice assistant users.
User Research
We recruited users from a broad demographic to learn how they would interact with a virtual assistant.
What We learned and Next Steps
After the first round of testing we learned that a voice version of existing app features is not the functionality our users are looking for with a virtual assistant. We also learned most men had an adverse reaction to the male assistant voice. What we learned was voice allowed users to navigate and dig deeper into their own spending habits which users seem to enjoy. The flatten of navigating the app was also another highlight we notice. Going into our next sprint we develop helpful assistant that can automate your spending and finances. We tested the prototype below.