Ask Backtrack

Overview

Backtrack helps professionals capture meeting notes, but users struggled to turn them into action. I designed an AI-powered assistant that transforms notes into follow-up emails, summaries, and task lists—balancing speed with user control. The challenge was making AI feel approachable and trustworthy, not robotic.

The Challenge

Research

I analyzed leading productivity tools to understand what works and what frustrates users.

Key Insights:

Design Process

Prompt Engineering

❌ Too vague: "Help me with my meeting notes"

❌ Too rigid: "Generate a 3-paragraph follow-up..."

✅ Just right: "Draft a follow-up email for today's call"

I tested multiple prompt structures to balance specificity with natural language. I used ChatGPT to generate realistic copy during design—not lorem ipsum—which helped test tone variations and identify edge cases early.

Iteration Process

Final Design

Key Features:

Conversational Design

Every word matters. Friendly, helpful language makes AI feel like a teammate, not a machine.
Outcomes

Outcomes & Key Takeaways

What Worked:

Pre-built prompts eliminated the "what should I ask?" problem. Tone controls built trust by giving users refinement power. Using ChatGPT for realistic mockup copy revealed edge cases (overly formal tone, response length issues) early.

Key Takeaways: