Designing with AI: Zoomcar Dashboard Exploration
Used AI tools like Claude, Bolt, and Figma Make to design Zoomcar’s vendor dashboard. Explored what works, what doesn’t, and how to design smarter.
App name / Client
Zoomcar
My Role
Product Designer
Industry
Automotive
Platform
Web
TL;DR — The AI Design Tool Scorecard
| Tool | What Worked | What Didn’t | Verdict |
|---|---|---|---|
| Claude | Got me unstuck. Good base layout ideas. | Needed lots of manual pruning. | Great for brainstorming |
| Stitch | Nudged simplicity. | Output wasn’t very usable. | Wasn’t very useful for this project |
| Bolt | Structurally sound. | Clunky, less intuitive for designers. | Great tool, heavily dependent on good prompting |
| Figma Make | Amazing UX. Intuitive layouts. Effortless prototyping. | Mostly nothing. | My first tool to go-to from now on |
The problem:
Zoomcar isn’t just for solo car owners anymore. It's grown to include large business partners. We are talking about hosts managing fleets of 200+ cars. Each of these organisations has multiple managers, each managing a slice of that fleet. But the current “My Cars” tab on the app just wasn’t built for that scale. Tasks like updating bookings or finding car details become frustratingly inefficient at fleet scale. What we needed was a smarter, faster way to manage lots of cars without pulling our hair out.
The Solution:
A vendor dashboard for the web. Think: data management at scale, batch operations, and an interface that stays clean even when the data gets messy.
To be fully honest, I have used this project to go full AI-nerd. I tested different text-to-design tools and documented everything. What worked, what sucked, what surprised me. This blog is a reflection of that learning journey.
Phase 1: Getting unstuck with Claude
I was only provided a vague business requirement doc initially. No direction, no design—just a list of features that needed to be on the dashboard. So I turned to Claude.
The prompt was essentially just the doc and some more context about the problem.
Claude's output:
Takeaway:
Claude gave me a decent skeletal structure, some good ideas, some bloated extras. But it was enough to get me started. AI's only as good as your prompt, and this made that very clear.
Building off Claude’s structure, I whipped up a first draft of the dashboard focused on **Bookings, Cars, and Pricing (**the core v0 functionality).
We ditched the homepage (not enough to show yet) and kept things minimal, with alerts and filters for fast nav. This version worked, but felt off still.
Phase 2: Enter Google’s Stitch
round this time, Google launched Stitch, their AI UI explorer. I tossed in my newly acquired PRD (finally) and waited for a better direction.
Result:
It didn’t blow my mind, but it did help push me towards a cleaner layout.
This was also when we introduced a key feature “Swap Car”. Instead of Zoomcar auto-assigning a backup car from another host, managers could now offer a replacement from their own fleet. Super useful.
Updated UI:
Phase 3: The search for a better form
After a backend convo, we dropped search for a filter form to reduce API calls. This form would bundle all filtering into one clean action, with bonus features like multi-car license input.
To prototype this, I used Bolt and learnt how to prompt engineer via ChatGPT (I tried to squeeze into this project as many things I could try my hands on. Find the prompt engineering prompt here).
Verdict:
Bolt got the job done. The structure was usable, though not super designer-friendly. Still, it helped me arrive at our final v0 draft:
Phase 4: Special appearance ft. Figma Make
Figma finally released Make in beta. This was the tool I was desperately waiting for. It uses Claude Sonnet under the hood but feels way more tailored for designers (finally, someone thought of us first and not just devs).
Using the same prompt I gave Bolt, here’s what Figma Make gave me:
Why I loved it:
- Intuitive designs
- Cleaner UX decisions
- More hygiene-level elements by default
- Built-in prototyping that just works
From here on, I fully adopted Figma Make for the project and kept evolving the prototype for v1.
It also generated extended flows, like this side panel for car details + pause booking functionality, with zero struggle:
🔗 Full prototype exploration (built in Figma Make)
Final Thoughts
This isn’t a dig at any one AI tool. It’s not a “which one’s better” showdown—it’s a curiosity-fueled field test. And what I’ve learned is this: different AI tools shine in different contexts. Claude gave me a solid jumpstart. Bolt helped with structure. Figma Make felt like a creative partner. Even the less helpful tools taught me something.
The truth is, there’s no perfect one-size-fits-all solution. But that’s what makes this moment so exciting.
So keep experimenting. Keep prompting. Keep pushing buttons just to see what happens.
It’s a wild, adventurous time to be building, and if you’ve got even a little spark of curiosity, now’s the time to let it run wild.
You can email me @ shivanisingh54996@gmail.com or reach out to me on X / LinkedIn