I use AI every day designing products for millions of people. Here’s what actually separates the designers who get it from the ones who don’t.

1. Prototype with AI — don’t just design with it

Stop presenting Figma screens to stakeholders.

Build something that works instead. When I switched to functional prototypes, people stopped debating button colors and started reacting to whether it actually solved their problem.

15 minutes. Full alignment. Done.

Start here: Use Cursor, VScode, Builder.io, Claude or Figma make. Describe the flow. Direct the build. You’re not coding — you’re making decisions.

2. Know what to feed it

This is the real skill. Not using AI — feeding it the right context.

Bad prompt: “Design an onboarding flow for a fintech app.”

Good prompt: “Design an onboarding flow for first-generation immigrants opening a US bank account for the first time. Skeptical of banks. Unfamiliar with financial jargon. On mobile plan with limited data. Trust matters more than speed.”

Same tool. Completely different output.

Here’s what AI can never do — go to where your users are (This is your super power). Watch them struggle. Hear the thing they don’t say out loud. That thinking, that judgment, that firsthand knowledge is what you bring to every prompt. It’s the one input AI can never generate for itself.

3. Use it to find what you missed

After you design the main flow — ask AI to complete everything you skipped.

Error states. Empty states. Edge cases you quietly noted and moved on from.

Prompt: “What happens when the user has no data, enters wrong information, or takes an unexpected path? Build those states.”

It won’t replace your thinking. It’ll reveal the gaps in it.

4. Prompt like a designer, not a user

Most people prompt like they’re Googling something.

Designers should prompt like they’re writing a brief.

Audience. Constraint. Tone. Outcome. All of it upfront.

Bad: “Write an error message.”

Good: “Write three error messages for a failed payment. User is mid-transaction, frustrated. Tone: calm, solution-focused. Under 20 words. Include a next action. Also provide a version under 40 characters for mobile — ensure it doesn’t break the component.”

That last line matters. Designers think in constraints. Most people prompting AI don’t. That’s the difference.

5. Build visual systems, not one-off images

Generating one great screen is easy.

Generating 20 that feel like the same brand — same palette, same style, same level of detail — that’s the skill.

Before you generate anything, build design system, build components just like you would in Figma. Any good AI design tool Figma make, Claude support it.

6. Build a personal knowledge engine for every project

Discovery done. Interviews done. Transcripts everywhere. Nobody has time to document any of it.

Feed everything into AI as you go — meeting notes, transcripts, research findings, your own observations. Then use it as a search engine for your own project.

“What did users say about the checkout flow in week 2?” “What patterns came up across all five interviews?”

Instant answers. You stop rebuilding context and start building on it.

Use that knowledge base to structure your next presentation, sharpen your POV, write your design rationale.

The AI didn’t do the thinking. It made sure none of your thinking got lost.

7. Know when to put it down

This is the one nobody talks about.

AI is exceptional at generating. It cannot decide what should be generated.

Which problem is worth solving? Which feature should exist? Which direction actually serves the user?

That requires being in the room. Reading the hesitation. Knowing why three previous solutions failed.

AI handles the making. You handle the deciding.

In a world where everyone has the same tools — the designer who decides better wins.