For this project, I knew it was going to be a challenge to get on the phone with customers of Scout Fuel, so I leaned into AI to help get situated in the needs of the users I was designing for.

My prompt was to get Claude to tell me the story of a Fuel Manager at a trucking company to help me understand their jobs-to-be-done and general challenges.

I then prompted a narrative vision of what an ideal product might unlock for a Fuel Manager. I did this in narrative form, as I have found that stories are often the best way to communicate and to capture the end result.

From here, I set out to design and build a product that was supportive of this vision.

Because my goal was to build a front-end prototype they could continue building on, I jumped straight into code and used the Shadcn component system as the base of the prototype.

Shadcn gave me modern componentry, and then I used Tweak CN to customize three unique style directions for Scout Fuel to review and provide feedback on. That gave us a solid design foundation to start building out the actual experience.

I also used Tailwind for layout and styling, so the stack stayed squarely in today's React ecosystem: composable UI primitives, utility-first CSS, and a codebase pattern product engineering already knows how to run with.

With a clear vision set and the scaffolding of the design foundation in place, I used Claude to shape the first one-shot prompt and get a V1 on the canvas. From there, I moved into Cursor as my primary environment for both development and design, and the work of refining kicked in.

Once AI put the first version on the canvas, the work shifted to subtraction. The first pass usually has too much: extra panels, extra lines, extra decoration.

I start by moving things around, merging repeated blocks, and removing anything that doesn't carry real signal. It's less about polishing and more about reshaping the structure so the core workflow reads quickly.

That phase feels more like carving than layering. Each pass strips away noise so the product gets closer to the rough shape I have in mind.

After the refinement phase removed distractions and clarified the core shape of the product, I started adding new features focused on engagement so the experience wasn't just a system of record.

I introduced a dynamic efficiency score based on each trucking company's purchase history; the score moved up or down with fleet driver execution and became a prominent dashboard signal for quickly identifying which drivers needed attention. The goal was to bring a grounded sense of gamification into the workflow.

That sequencing matters. Gamification without structure reads as desperation. Structure without warmth reads as software that doesn't respect long shifts.

In the repo I added a Cursor skill that summarizes every pull request and appends to that Notion doc. I told the agent to write in customer-facing language — what value landed for Scout Fuel, not just commit noise — and to capture screenshots of the work so the log stayed visual, not abstract.

After each session building the prototype, I would type “summarize” and let the skill run. It turned into one of the highest-leverage habits of the sprint: a lightweight audit trail that made it much easier to communicate design decisions back to stakeholders without rebuilding context from memory.