Context and Objectives
The development of Ask Robin was a long-awaited step in enhancing the platform’s analytical capabilities. While an initial design concept existed, it was never implemented due to shifting priorities. However, with growing demand from both existing and prospective customers, the company decided to push forward with an MVP, with a clear focus on delivering a solid foundation for future iterations.
My role was to design this MVP, balancing efficiency, usability, and scalability while collaborating closely with the product manager and development team.
MVP Scope & Requirements
For the initial release, the product team identified six core functionalities that needed to be included:
Creating a new conversation – Users should be able to start fresh discussions with Robin at any time.
Opening and continuing past conversations – Users should navigate their conversation history and pick up where they left off.
Deleting a conversation – Users should have the ability to remove past discussions when needed.
Renaming conversations – Automatically generated titles should be editable to enhance organization.
Editing the last query – Users should be able to modify their most recent question, with a clear indication that doing so will generate a new response and overwrite the previous one.
Providing feedback on responses – Users should be able to rate answers as helpful or unhelpful, with structured feedback options for unhelpful responses.
With these functional goals in place, the next step was to translate them into a streamlined user experience.
Designing the Core Interactions
1. Starting a New Conversation
To ensure easy access, the "Start New Chat" button was positioned in the history panel. This button is active when viewing an existing conversation and disabled when already in a new one, preventing unnecessary duplication.
2. Navigating and Continuing Past Conversations
A structured side panel lists all past conversations, categorized by time (e.g., "This Month," "Last Month"). Clicking on any conversation reopens it, allowing users to continue where they left off. The panel itself is collapsible for a distraction-free experience.
3. Managing Conversations (Delete & Rename)
Users can delete conversations via a contextual menu, with a confirmation step ensuring they understand the action is irreversible. Similarly, renaming is enabled through inline editing, overriding the system-generated titles.
4. Editing the Last Query
To allow corrections without creating redundant conversations, users can edit their last message. A tooltip and inline edit mode guide them through the process. Since modifying the query generates a new response, the previous one is temporarily disabled, and a confirmation popup clarifies the change’s impact.
5. Feedback Mechanism for AI Responses
Every AI-generated response includes thumbs-up and thumbs-down icons. If a response is marked as unhelpful, a structured feedback form appears, allowing users to categorize the issue (e.g., inaccurate, outdated, irrelevant). This ensures qualitative data collection without forcing users to type unnecessary details.
Designing the Core Interactions
Loading and error
Through collaboration with the development team, we identified the need for UI states covering system failures. Two key scenarios were addressed:
Loading State – To indicate that Robin is processing a query, a placeholder animation ensures users are aware of ongoing activity.
Error Handling – If the system fails to generate a response, a structured error message is displayed with a retry button, keeping users informed and guiding them toward the next step.
Looking Ahead: Introducing Graphic Responses
As development progressed, the product team decided to enhance Robin’s responses beyond plain text. The next phase of design included:
Data Visualizations – Charts and tables were introduced to make financial insights clearer and more digestible.
Flexible Layouts – The chat interface was adapted to accommodate different response types while maintaining consistency in usability.
These enhancements set the stage for a more interactive and informative AI assistant.
