CASE STUDY
Designing an AI-Powered Insights Feature Using a Design Sprint
Overview
As the lead product designer on this project, I helped design an AI feature that enables users to get meaningful answers based on the data within their accounts. The goal was to integrate this capability in a way that was intuitive, trustworthy, and immediately valuable to our users.
Problem
Users needed a faster and easier way to extract insights from the complex datasets within their accounts, without relying on external tools or manual queries. Our challenge: design a solution that empowers users to ask natural language questions and receive relevant, accurate responses based on their own data.
Approach
The Design Sprint: To accelerate alignment and exploration, we conducted a 5-day design sprint following the methodology from Jake Knapp’s book, Sprint.
My Role
- Lead UX designer in design sprint, sketching sessions and user testing
- Created and iterated on the high-fidelity Figma prototype
- Conducted user interviews and synthesized findings to help validate feature value
- Collaborated with PM and engineers to scope MVP
Day 1: Understand & Map
1. Defined our long-term goal: Help users get trustworthy answers from their own data through a conversational interface.
2. Key sprint questions:
- Can users trust the AI's responses?
- Will users know what to ask?
- Will the AI create useful, accurate experiences?
- Will people use our software in a totally new way than they've used web-based software for decades?
3. Created a simple map of the AI workflow, from asking a question to seeing results.
Day 2: Sketch
1. Sketched multiple UI/UX solutions individually.
2. Included different approaches to: Input experience, results display, and user feedback on response accuracy.
Below are my sketches that I presented to the team.
Days 3 & 4: Decide and Prototype
Day 3: Decide
- Reviewed sketches as a team.
- Voted on the most promising ideas.
- Combined the strongest elements into a single prototype concept.
Day 4: Prototype
- Assisted in building a realistic, high-fidelity prototype in Figma.
- Focused on clarity, data confidence cues, and user onboarding into the AI feature.
Days 5: Test
1. Conducted 1:1 usability testing with five existing users of our software.
2. Collected qualitative feedback on:
- Ease of use
- Usefulness of responses
- Trust in the AI's capabilities
- Asking the value they see in this AI feature? Would you pay more for it?
Results & Key Insights
- Users appreciated the natural-language interface but requested guidance on what types of questions to ask
- Data confidence labels and source links were critical in building trust
- Some users wanted a "recently asked questions" history for reference and reusability
MVP Development
- Added question suggestions on launch
- Included a confidence score and data source link for each answer
- Enabled saving and pinning questions for future use
Impact
- Over 70% of testers said the feature helped them find answers faster than existing workflow
- The design sprint accelerated alignment and helped de-risk investment in AI by validating value early