<aside> ⚠️

Disclaimer: This case study is 100% AI simulated

Tool used: ChatGPT, Hotjar, Dovetail, Figma Make, Figma Slide

</aside>

Stakeholder → jump here

UX Story Insight → jump here

Problem Framing

“Users struggle to confidently select the most suitable ChatGPT model due to limited knowledge, unclear guidance, and a lack of behavioral transparency. This results in suboptimal outcomes, rework, and hesitation in high-value workflows.”

Key Findings Summary (from User Interviews)

1. Confidence & Expertise Levels

• Majority of users lack confidence when choosing a model.
• Only 2 out of 5 users felt confident (rated 4+); 3 users were uncertain (rated 3 or lower).
• Some users didn’t even know model selection was an option or necessary.
• Simulated metric: ~60% of users are unaware of model selection, and incorrect choices lead to an average of 8 minutes lost per task.

Implication: Users need support, clarity, and just-in-time education to confidently select the right model.

2. Current Selection Criteria

• Most-used factors:
  • Accuracy / reliability
  • Response speed
• However, default bias is high: users often stick to the default because they don’t understand the differences.
• Simulated metric: 70% of users default to GPT-3.5 regardless of task type, missing opportunities to leverage model strengths.