Startup

SXSW

Conversation design, strategy, & product management of SXSW’s in-app & Facebook Messenger virtual assistant

Conversational Strategy

As a part of the Voxable team, I helped SXSW develop their conversational intelligence Abby and helped manage the chatbot development that helped attendees explore SXSW events, artists, venues, and information via the SXSW app and Facebook Messenger.

Attending SXSW is an overwhelming experience, and most people need help accessing and managing all the information about events they want to attend and places they want to go. I designed Abby to help attendees intuitively find the information they need without requiring additional support from SXSW staff. I helped the SXSW technology team understand how a conversational assistant would help attendees navigate over 6,000 events across 600 venues.

Conversation Design

I audited the existing SXSW communication points across their social, customer service, and applications and discussed what conversations and tasks could be automated with the team. I then analyzed the thousands of data points that describe the SXSW event and all its moving pieces. I used the existing communication and data to design a chatbot information architecture that allowed users to easily browse events, artists, and venues and ask questions about the event. 

I crafted Abby’s dialogues with users and iterated upon the voice and tone established by the SXSW brand team. I designed the conversational flows and mapped them all out in diagrams that informed development of Abby. 

I developed a messaging UI language based on the SXSW app and Facebook Messenger’s messaging UI to deliver high-quality content to attendees.

A demo of the natural language conversations user can have with Abby in Facebook Messenger.

Product Management

I helped train and test Abby’s AI systems and manage the delivery of the Abby production app. After mapping how users flow through the numerous conversations they can have with Abby, I worked with the developer to build the natural language understanding (NLU) model that translated attendees’ text messages into the structured data the intelligence used to deliver a response.

We developed the SXSW intelligence in Voxable’s open-source CUI framework, Heliogram, which we designed to integrate with an app’s NLU model as well as incorporate business and UX logic.

Thousands of attendees reached out to Abby on Facebook Messenger, and the automated assistant achieved an impressive accuracy rate. Our solution outperformed previous solutions by being able to carry on deeper conversations with users and answer their questions better. We successfully prepared our codebase for the scale of the massive event in which 432,500 attendees descended on Austin throughout the ten days of official SXSW programming.

Check out the SXSW post when Abby debuted in 2018 ↗︎