Watson Conversation lets users quickly build, deploy and maintain a chat bot.

My Role: Design Lead

the challenge

People spend more time using messaging apps than any other app and businesses wanted to connect with those customers, without making them switch apps and contexts. With the business need growing, how could we leverage the power of IBM's Watson platform to give them a way to easily connect with their customers?  

We started by organizing a three-day workshop to align design, development and engineering teams. From this initial meeting we defined our users, created and prioritized user stories and sketched storyboards for a first experience.

Sketches, Storyboards and Wireframes

Through rounds of iteration based on the workshop, I sketched out storyboards to gain a rough understanding of the story we wanted to tell. Once we felt comfortable with those I translated them to digital using Sketch and then created more detailed wireframes of our initial area of focus, letting the user easily build their conversation using a dialog tree approach.

User Journey: Build > Deploy > Improve

For the initial release, we wanted to focus on giving the user the best experience to build their bot. We also identified two other experiences that we prioritized quickly after. The Deploy area let the user quickly release their bot to the world, and Improve gave direction on what users could do with the data once people start interacting with their bot.

Build

We chose to represent the building process of a chatbot as a dialog tree. We found that this was the most effective visual way to communicate how the system was processing through a user input. This also helped users build out a visual architecture they could become familiar with while building their bot, and see the shape of their data evolving over time. 

Deploy

When we launched the UI, deploying a bot required a lot of code, but our target user was not comfortable with the level of technical knowledge needed to do this task. Providing easy and quick deployment options was critical to help users put their bot out into the world, without needing to write code. 

Improve

Once the user released their bot, all of the interactions would appear in this section. Based on that data we were able to craft a dashboard of statistics that were meaningful and actionable. When a user went to view the activity logs, we also gave them actions to take that would help them train their machine learning system. This user information proved to be the best and fastest way for our users to continually improve their bot’s performance.

Try it out

We also discovered that being able to test the dialog in real time was extremely important. We designed a piece of the UI that allowed a user to play the role of their end customer and chat with the dialog they just built. We also found that the ability to test their build was needed at each stage of the process and not just while they were in the Build phase. Knowing these, we chose to make the testing panel accessible through every section of the UI.

Customer outcomes