AWS QUICK SUITE • Concept 2025
Build with AI
My Role
UX Design lead, Prototyping, Stakeholder buy in
Timeline
6 weeks, Launch Nov 2025
Context
Quick Automate is a tool that automates complex enterprise processes like claims processing, appointment scheduling, compliance checking. These workflows involve conditional branches, loops, and cross-departmental handoffs that are difficult to visualize and even harder to build without errors.
Problem
Traditionally, these automations are built by trained automation developers due to their complexity. The goal was to lower the barrier of entry to building so that business users like product managers can also build and deploy them.
The Solution: Chat-to-Workflow
A conversational interface to transform a technical chore into a simple dialogue.
Intent Recognition
Instead of mapping out 20 manual steps, a user simply describe what they want to do. The system interprets the goal and handles the background "plumbing."
Abstracting the Technicals
The chat interface hides the complexity of switching between UI and API. The user doesn't need to know how the data moves—only that it does.
Human-in-the-loop
If the AI agents encounter a vague step, they simply ask for clarification via chat, making the setup feel like a collaboration rather than a programming task.
Key UX Flows
Initiate with prompt or document
Users can start building automations by simply describing what they want to achieve in natural language or add process document, eliminating the overwhelmness of a blank canvas and 100s of actions.
Make edits
Through both the canvas and chat, help users easily locate the changes made by AI. By visually anchoring AI modifications in both interfaces, we eliminate 'black box' anxiety and empower users to review, refine, or revert automated actions with total confidence.
Suggested next steps
Contextual prompts recommending the next logical build step, guiding users through the automation creation process with intelligent suggestions.
Test
Contextual CTAs on the canvas to fix errors with AI (chat)
Version 1
Upon joining, I conducted an audit of the existing v1 designs to identify friction points and alignment gaps with our updated product goals
Insights
Our researcher conducted a qualitative study with solution architects and automation developers and focused the study on authoring UX flows which included both the chat experience as well as the canvas experience with dragging and dropping nodes. I leveraged the existing designs, established prior to my arrival on the team, as a prototype for initial user testing. I used those insights to drive the subsequent redesign.
1. Help from AI
Beyond just building the automation, users want help from the chat throughout the building process
2. Troubleshooting is a primary path
Users expended a significant amount of effort to resolve the errors in their automations with minimal success.
Design Tenets
I defined a set of foundational tenets that acted as a guiding direction for our cross functional team to make decisions. These principles ensured that while the AI provided speed and "magic," the user remained the ultimate authority in the development loop.
Intuitive Navigation
Enable users to find features and navigate the tool using natural language. This reduces the learning curve by letting the chat act as a guide to the right capabilities.
Streamlined Workflows
Automate repetitive, multi-click tasks to save time. The AI assistant implements changes and surfaces relevant components, maintaining transparency by informing the user of every action taken.
Human-in-the-Loop Design
Since Gen AI can err, we prioritize iterative refinement. Users can review, accept, reject, or modify all AI suggestions, ensuring they maintain ultimate control over the automation.
Familiar yet Contextual
Incorporate the conversational patterns users expect from tools like ChatGPT, while ensuring the AI remains deeply aware of the specific automation development environment.
Parity & Control
Maintain a 1:1 feature parity between the chat and the canvas. Every action available via AI must also be accessible manually, ensuring no functionality is "locked" behind a conversational interface.
Interface Breakdown
AI Response
Natural language answers plus a summary of canvas changes to ensure transparency and easy verification of AI edits.
Action Buttons
One-click manual overrides (e.g., Accept/Reject) for immediate control over AI modifications. Clears on new input.
Suggested Actions
Contextual prompts recommending the next logical build step. Transitions from static (GA) to dynamic (post-GA) logic.
Status Indicator
Real-time system feedback (e.g., "Test Running") based on current canvas activity. Clears on new input.
Step Shortcuts
Persistent navigation tools for the selected step, providing quick access to details and neighboring elements.
Outcome
This work was launched in Quick Suite in November 2025. Since the launch, there's been consistent usage of the chat feature to build automations, though more work is underway to make the chat experience useful and intuitive for non technical users.