How to Become an AI Designer: A Practical Guide to Shipping Design Yourself
1. Move away from Figma-centered work and validate the flow quickly
To keep up with development speed, change the feedback loop before pixel perfection.
- Close Figma for a moment and use Claude to create clickable HTML prototypes or exploratory UI components.
- Repeating short cycles of sharing within a day, getting feedback, and remaking them quickly makes complex user flows concrete.
- There’s no need to reproduce every screen with the same level of polish as the actual product.
- Choose lo-fi when you want to understand structure and flow, and hi-fi when you want to check visual polish, and separate that from the cost of the last 10% of detail.
2. Move the point of reference for design from Figma to the actual product
In an environment where implementation speed outpaces prototypes, the shipped product becomes the latest design document.
- A prototype becomes less like a deliverable for conveying finished screens and more like a tool for adjusting the direction of implementation in real time.
- When designers have access to the codebase, they can directly refine qualities that are easily lost in handoff, such as spacing, color, responsive behavior, and interactions.
- This is where using AI to make work faster and using AI to make the actual result better diverge.
3. Expand from small PRs to frontend ownership
Rather than developing the entire frontend, expand into a role that takes responsibility for key decisions in the experience visible to users.
- Start by shipping small changes like copy, color, and spacing, and find an engineer champion who can help with initial setup and code reviews.
- If code access is difficult, you can experience your first PR through a lower-barrier method like Cursor Bot in Messenger.
- If it’s hard to wait for organizational support, you can start a personal project with v0 or Lovable, then broaden the scope of work with Cursor or Claude Code.
- It’s important to build trust gradually, starting with small UI tweaks and extending to UX improvements and frontend decision-making.
4. Accept the confusion that comes with faster speed
Running multiple agents and submitting several PRs a day increases productivity, but it does not guarantee satisfaction with the creative process.
- Working by constantly moving between tabs, prompts, and PRs creates a new kind of context switching and fatigue.
- It is too efficient to go back to the old way, but there remains a tension that is not satisfying enough to stay the same.
- The core of being an AI designer is less about the ability to generate more and more about the sense for judging and protecting good design even as the production process speeds up.
