Product design
AI product
This tool's goal was to enable every line of business at Wells Fargo—from retail to home lending to internal HR—to create a custom chatbot. The user base was our internal conversational designers, who would built and train the chatbots to improve their performance. This was pre-LLMs, so it was a laborious process that involved huge trees of potential "paths" a user could take. Dealing with this scale of data was a major challenge.
I was one of two designers on this project, along with a team of engineers and a PM. Together, we worked on a UI that supported long tree structures: the top level was called a "response flow," which was comprised of blocks that could link to many other blocks. Within each block, users could also add custom slots that allowed for variable data.
If the chatbot detected a user's "intent," or goal of engaging with the chatbot, the user could be given different responses or asked for more information. The conversational designers had a great deal of flexibility, allowing them to pivot as they learned more about how users engaged with the chatbots.
We successfully released the new version, enabling more full-scale conversations; we also migrated use cases over from the previous generation without disruption, a major point of concern when dealing with a user base as large and wide as Wells Fargo's.
