When I first heard about AI-powered design at Config this year in San Francisco, I was genuinely excited. It felt like the beginning of a new era in how we approach design, especially with the rise of AI-driven experiences and new AI prototyping tools. Since then, platforms like Lovable, Galileo AI, Uizard, and Vercel’s v0 have been making waves, opening up possibilities that seemed far-fetched just a few years ago. As designers, we read countless articles, watch demos, and stay curious about what these tools can do.
When I finally decided to try Figma Make for myself, the excitement was real. But my very first attempts before learning any best practices or understanding how to truly guide the AI were far from what I had envisioned. The outputs felt incomplete, and I quickly realized that reaching a final, polished product on the very first try is incredibly difficult with AI tools.
Just like in a traditional design process where we progress step by step, get to know the user, conduct research, study design systems, identify pain points, and only then start designing, working with AI requires the same mindset. We need to learn how to express our intent in the “design language” of our prompts so that the AI understands it as clearly as a human collaborator would.
This is not just about Figma Make. When you think about the direction of the industry with tools like Lovable, Galileo AI, Uizard, Vercel’s v0, and others, it is clear that design is shifting toward AI-enhanced workflows. In this era, designers must develop the skill of writing precise prompts to accelerate the design process and use AI to improve both speed and quality of outcomes.
As the saying goes, “AI will not replace designers, but designers who use AI well will stand out.”
This skill is often referred to as Prompt Engineering. In many ways, writing a good prompt is like writing a strong design brief. It is about providing the right context, details, and constraints so the AI can deliver exactly what you need.
Mastering Prompting as a Product Designer
Now that it is clear that prompting has become a core skill in a modern Product designer’s toolkit, the real question is: how do we take this skill from basic to exceptional? Like any design discipline, it requires method, structure, and practice. Let’s break down the principles that can make your prompts far more effective and the results much closer to your vision.
1. Avoid starting with a generic prompt
Writing a generic prompt like “Make a modern event app” will rarely get you to the exact result you have in mind. Sure, it might give you a starting point or a rough idea, but it won’t match your vision. AI can’t read your mind, so you need to guide it with a well-structured prompt that clearly communicates what you want.
Result: The output did include a few elements that sparked ideas, but it was still far from the design I had envisioned. It lacked the context, priorities, and constraints we would never skip in a traditional design brief.
2. Use a structured prompt
Once I started using the structured format suggested by Nick Babich, the quality of my AI-generated screens improved dramatically. Instead of throwing vague, one-line prompts at the tool, I broke them down into clear, specific sections. This ensured the AI understood not only what I wanted to design, but how it should look, feel, and behave.
Here’s the structure I follow:
This is the detailed, structured prompt I created for my event app concept, followed by the resulting UI:
3. Use Buzzwords strategically
As highlighted in Lovable’s Buzzword Hacks guide, using specific design jargon like “elegant gradients” or “glassmorphic cards” can give AI very clear visual directions. These terms work like shortcuts, helping the model instantly understand the style you are aiming for.
For example:
“Minimalistic UI with glassmorphic cards and soft drop shadows”
“Dynamic hero section with parallax scrolling effect”
When used intentionally, buzzwords can bridge the gap between your idea and the AI’s interpretation, speeding up the path to a more accurate first draft.
4. Speak the Design System language
Be precise in your prompts by referencing design system rules and layout standards. This helps AI generate designs that are consistent and easier to maintain.
Examples:
“Use 12-column responsive grid with 32px gaps”
“Apply consistent vertical rhythm with 64px section spacing”
“Mobile-first, responsive layout”
If you’re including a Figma file as a reference:
Ensure auto-layout is correctly set up.
Use meaningful layer names for clarity.
Include your component library in the prompt so AI applies your brand styles and components consistently.
5. Provide a UI reference
You can upload a screenshot of a product you like and simply tell the AI, “Match this style.” Alternatively, you can create a rough wireframe in Figma and use it as a reference image.
In my own experiments, I noticed that adding a visual reference made the results feel much closer to what I had in mind. It gave the AI a concrete visual foundation to work from, reducing guesswork and aligning the outcome with my vision
6. Develop your prompt with ChatGPT or Claude
Instead of struggling to write a perfect prompt entirely on your own, use AI to help you shape it.
Tell it your needs, your target audience, and your preferred style, and let it return a structured, detailed prompt you can then refine further.
7. Break down the design process
Before building full user flows, first lock in the visual language. Decide the look and feel, establish the style, and get the core layout right. Once the design direction is clear, move on to functionality, motion, and behavior. This sequence produces cleaner, more consistent results and reduces rework.
Example iteration plan
First iteration: design components only (cards, chips, buttons, inputs) with spacing, states, and typography.
Second iteration: compose page/screen using those components and finalize layout, hierarchy, and copy.
Final iteration: add animations and interactions (entrances, transitions, micro‑interactions, loading) and tune easing, duration, and stagger.
8. Iteration mindset
Think of Figma Make as a brainstorming tool, not a magic “one-shot” designer. Don’t expect perfection on the first try.
Instead:
Refine: Adjust prompts based on what’s missing or unclear.
Critique: Evaluate the output critically, noting both strengths and gaps.
Reprompt: Feed those observations back into the prompt to get closer to your vision.
Each cycle should bring the result incrementally closer to what you have in mind.
9. Use real content
Replace generic placeholders with realistic data like actual dates, prices, and user names to make your designs feel authentic.
Ask Figma Make to show the type of data you expect, add a “data import” option for control, and start with simulated data to avoid API risks.
Extra tip: Figma Make can use your device’s inputs (keyboard, audio, camera) for richer prototypes.
10. Editing techniques
Point and Edit – Make quick visual tweaks directly on the canvas.
Access the Code – Adjust small code snippets to see instant results.
Screenshot + Custom Prompt – Show the issue visually and describe it in your prompt for a targeted fix.
11. Use version control
Save a version after each iteration: “Save this version as Version 97, added hero image and changed CTA text.” Write your next prompts on top of this saved version.
This has been one of the most useful techniques for me. 🤓 If you don’t save, some prompts can unexpectedly change the entire page. Be specific about what needs to stay the same, and always store the version in case you need to revert to it later.
12. Animation best practices
Be familiar with core animation terms (fade in, slide up, ease-in-out, spring).
Describe the flow to ChatGPT: “Cards fade in sequentially from bottom with 150ms stagger.”
Maintain consistency: Ensure your components, text, interactions, and animation language all follow the same style.
This is a journey. We are all learning, and the best way to learn is by experimenting. Share what you try and help strengthen the community.
I will also share my own learnings in my upcoming newsletters. If you would like to stay updated, you can subscribe. 🚀
Thanks for sharing!
Great!! Thanks