I’ve always been interested in AI and machine learning, so moving into conversation design seemed like a lateral one for me. Chase’s virtual assistant launched in mid-2018, so it’s still a very new product. For a year and a half, the team focused on voice activation, but that feature was de-scoped in January 2020. The content was previously tailored to be said aloud to our users, and I’m re-stitching it into more suitable chatbot material. I’m refining the voice to be more natural and the tones to be responsive, varying between professional, personable, and gently self-aware that it’s a chatbot.

As I came aboard the project early February 2020, I still have so much to learn about this fairly new frontier. I’m enjoying the technical aspect of mapping strings to relevant tokens with our NLU (natural language understanding) tools. I’m going about this with design best practices, collaborating with product and design as early as possible for the best, most up-to-date prototypes.

Because this was supposed to be a voice assistant, we’re scrapping and reworking a lot of our content. Brevity is not only key, but king, for chatbots. We’re not yet ready for official user testing, but we can still work with some old voice activation research results. Our colleagues have also given us valuable input during hallway testing. Virtual assistant’s audience is anyone that has the Chase app, so we’re pulling personas from other mobile-focused product teams.

See the examples of my work process: rewriting language with my rationale, a conversation map, and wireframes.