My task was to take the approachability and simplicity of Roku's TV products and translate them into an equally appealing voice UX. Both Roku and I were relative newbies when it came to the development of voice-driven software. I was there for six months, and during that time I became the go-to person for Voice UX. Roku announced its Entertainment Assistant at the beginning of 2018.
There was a lot to learn about voice - I read papers and watched presentations. I sat in a room and tested other voice-driven products. I quickly developed an opinion about what worked well, and what didn't.
I leveraged a small set of existing UX use cases, and broadened them across entertainment types. In the process I developed annotations that summarized the type of request a user might make. That made it easy to map use cases to Intents.
I tracked down the developers who would answer my questions about how our system was going to work, and then mapped out the UX components. In the end I was proficient enough to suggest architectural features that would benefit the UX.
The UX group had already developed a robust set of design principles for screen-based Roku products. I expanded the set to include Voice. These new principles were reviewed and refined with the participation of the entire UX team.