ChatGPT Degree Planner

The first Figma Make prototype explored a data-rich approach to degree planning, bringing many parts of the problem into one experience. Pulling it apart with the team surfaced promising ideas and, more importantly, several different ways the product itself could work.

I used those ideas as a jumping-off point to build three parallel concepts in Figma Make. Each explored a different balance between conversation and structured UI, density and utility, guidance and engagement, so we could compare working product models side by side.

That comparison helped us focus the decision around two questions: how much structure does an AI planning experience actually need, and how much playfulness helps people make it through a complex planning process?

The strongest direction used progressive disclosure. Instead of asking students to understand the whole degree-planning problem up front, the planner asks one question at a time: degree level, major, weekly study time, budget, timeline, learning style and school preferences.

Complexity appears only when it becomes relevant. The conversation carries the questions and reasoning, while visible structure gives students a way to understand and compare the plan, its cost, timeline and possible adjustments.

The structured steps provide a clear path, but they aren’t the only way through. Students can answer in the interface or use the chat itself to respond, ask questions, and reuse information they’ve already shared with ChatGPT. That lets the experience move faster when the context already exists instead of making someone enter the same information again.

We used a hosted proof of concept to test the direction with leadership and build alignment around the approach.

Dogfooding helped simplify both the questions and the answers. We shortened and combined the onboarding into nine steps, and changed how we communicated whether a plan was achievable. Instead of a “Realism” score, the final product describes the plan as a percentage “Realistic,” framing it as a current assessment users can improve through recommendations and adjustments.

Other ideas didn’t earn their complexity. A separate Burnout score could feel confusing or demoralizing, while having children and pets add flat percentages of “stress” made the model more playful without making the guidance more useful. We cut both from V1.