· 1 min read
Designers should care more about system design
AI helped me start building tools. System design helped me stop rebuilding them every week.
By Gaurav Chopra
Coming from a design background, I used to think mostly in screens.
How should this look?
How should this flow?
How should this feel?
But once I started building AI tools, I realised the actual product is often happening behind the UI.
How prompts are structured.
How memory flows between steps.
How outputs get validated.
How tools talk to each other.
How one small failure breaks the entire experience.
That’s system design.
And honestly, I ignored it at first.
I would build fast, duct tape things together, make it work somehow and for demos, it looked great. But after a few users, things started falling apart. Outputs became inconsistent. Logic became messy. Fixing one thing broke another.
That’s when I realised:
Good AI products are not just designed visually. They are designed structurally.
Now whenever I build something, I think less about “screens” and more about “systems.” That's what I did when I was building AutonautOS in Cars24.
What happens before the button click?
What happens after the AI responds?
What happens when the response is bad?
Can this scale beyond me manually fixing things?
As designers, we’re entering a phase where understanding systems might become as important as understanding typography or layouts.
Not because we need to become hardcore engineers.
But because AI products are deeply connected experiences. The magic is rarely in one screen. It’s in the orchestration behind it.