Amazon Lex
visual conversation designer
Creating an entirely new visual design tool for crafting chatbot conversation flows
ROLE
Lead designer
UX strategy
Information architecture
Design system contributor
DOMAIN
Desktop/web
Enterprise tools
Developer tools
Low code / no code builder tools
IMPACT
Reinvented Lex as a modern and more competitive conversation design tool
Made Lex more accessible to less technical customers
New interaction paradigms enabled easier creation of more flexible conversation flows
Updated UX patterns set foundations for future strategies for low / no code tools for AWS AI services
WORK ＞ AMAZON LEX VISUAL CONVERSATION DESIGNER
Context
Picking up from the v2 console redesign;
The next step in evolving Amazon Lex to a modern and competitive conversation design tool was to reimagine the experience as a non linear, visual canvas based design tool that would appeal to less technical customers, not necessarily familiar with a console environment.
Amazon Lex V2 console - Intent editor
Amazon Lex V2 console - Intent editor
Challenges / questions
What would I design if not constrained by the typical AWS console service design?
Should this be a separate product offering or integrated into the main console experience?
Who is really this for? Who are conversation designers and what do they need?
User goals
Quickly and easily get started building the conversation flow outline, without worrying about configuration.
Keep all the power and depth of Amazon Lex capabilities available as and when needed.
Maintain consistency with data and terminology in the console in order to effectively collaborate during the lifecycle of the chatbot building process.
Research insights
Understanding who the user is and what their needs are:
“Conversation designers are rare to find but at the same time bot developers do not understand the ‘ins and outs’ of dialog design ”
“…we have solutions consultants who design use cases, conversation designers who flesh out the dialog for each of these use cases, and then technical solutions managers (conversation architects) who are well versed with API integrations ”
Explorations [1]
Top to bottom / vertical flow direction.
Explorations [2]
Left to right / horizontal flow direction.
Connecting blocks
Research insights
Research conducted interviews with eleven conversation designers, voice user interface designers, and business users to discuss early concepts for the conversation designer.
How does this map to the console?
Existing users expected to be able to switch between modalities with consistent terminology and content organization.
I want to manipulate what I see
People wanted to be able to quickly just edit content inline without opening a full editor.
How does this thing scale?
People were concerned about how the visualization of more complex conversation flows would scale
They wanted to be able to switch from high level ‘scanning’ to low level detail editing
Hey wouldn’t it be cool if…?
People imagined being able visualize where they are in the flow, while testing
“I think they (conversation architects) would find value in this one tool that accommodates everything - just that at that point you must study the behavioral change that needs to happen”
“…they will not come in and start building the flow and everything like how you have it here - they will start with the boxes without adding the copy and just focus on what the conversation flow looks like. They will then come back and add the copy, utterances, training and other details.”
Final design
Canvas frame & tools
Configure blocks on canvas
Components & building blocks
Block configuration examples
Complete flow example
e.g booking a car
Outcomes and learnings
Delivered a new experience to meet ambiguous needs with strong collaboration from user research (the team was not used to working with UXR to synthesize generative research input to guide critical foundational design decisions).
Defined new more intuitive information architecture and workflows from scratch.
Repositioned the service as a modern and competitive option in the emerging field of conversation design tools.
Contributed useful and flexible new patterns to the design system.
Learnt a lot about the emerging field of conversation design and the people that do it!