Quick disclaimer: this post is more of a thought piece. I can not share the actual product or screens just yet, but I wanted to walk you through the process while it was still fresh. The approach is what matters here.
Yes, yet another article about AI. Shocking, I know.
But hear me out. This one is less about tools and more about process. Specifically, how I used AI to go from a meeting room full of ideas to a fully interactive MVP prototype in under 7 hours. If you work in product or web design, that number should get your attention.
Let’s get into it.
It All Starts in a Meeting
Like most things in product design, this started with a meeting. Stakeholders, the design team, and the development team all seated together to discuss a potential new feature. The goal was to align on scope, goals, and everything we needed to define before touching any design tool.
Once the direction was set, each team went off to brainstorm their own approach to the challenges ahead. But as always, design is where ideas start taking shape. My lead and I sat down, went through everything we had gathered, and once I had a clear picture of what we needed, the fun part began.
Stage 1: Context Dump and Research
My first step was to take everything from that meeting and dump it into Gemini. Not to ask for answers yet, just to provide context. I told it the full story of what we needed to build, the problem we were solving, and the experience we wanted to create.
From there I did some inspiration research, looking at other apps and tools with similar interactions and a look and feel close to what I had in mind. I pulled a few reference images and shared them with Gemini to help communicate the type of micro-interactions I was thinking about.
The key thing at this stage is that everything is still text and ideas. No design work yet. The goal is to get complete clarity before opening any design tool.
Prompt to use at this stage:
I am designing a new feature for [type of product]. Here is the full context: [describe the problem, the goal, the target user, and any constraints]. I will also share some reference images of interactions and visual styles I have in mind. Based on all of this, help me explore possible approaches, ask me questions if anything is unclear, and suggest any ideas I might have missed.
Stage 2: Building the Figma Make Prompt
Once I had all my ideas mapped out and had accepted a few additional suggestions from Gemini, I asked it to generate a prompt tailored specifically for Figma Make.
Why Figma Make and not Claude, Lovable, or another tool? A few reasons. We already have the company account, the design system is already set up inside Figma, and at this stage we are exploring and ideating, not building for production. No need to spend extra budget on additional tools when the right one is already in your hands.
Prompt to use at this stage:
Based on everything we have discussed, create a detailed prompt I can use inside Figma Make to generate a first interactive wireframe. The prompt should include: the purpose of the feature, the user flow step by step, the types of interactions needed, the tone and structure of any messaging, and references to the design system already in place. Leave some room for interpretation but be specific about the logic and flow.
Stage 3: First Iteration in Figma Make
I opened Figma Make, attached our design system library, and ran the prompt.
The first iteration honestly surprised me. It was much closer to what I had in mind than I expected. The preparation paid off. Having every stage thought through and clearly expressed in the prompt gave Figma Make enough to work with. After a few minor adjustments across about 10 versions, I had a fully interactive prototype ready to share for feedback.
The feature involved 6 questions with branching logic, different possible answers, and conditional messaging leading to different outcomes. Building that manually in Figma prototyping would have taken at least 4 hours just to get the interactions working at the same level. This first iteration took 1.5 hours.
At this stage we are not worrying about colors, typography, or visual polish. This is a low-fidelity wireframe that happens to work. That is the real value of AI-assisted ideation: getting something functional fast so you can test the experience and make real decisions.
Stage 4: Second and Third Iterations
After the first review I gathered solid feedback and moved into a second iteration with more detail and a clearer direction.
A heads up for anyone doing this: the more detail you try to control in Figma Make, the more back and forth you will have. Even with a design system attached, the tool has a tendency to swap out icons and make visual decisions on its own. That is a separate conversation, but worth knowing going in.
After 29 more iterations the second version was in a much better place. Still under 2 hours of total time. The third iteration took about 1 more hour for smaller refinements.
Total time from blank file to stakeholder-ready prototype: under 7 hours.
Prompt to use for iteration:
Here is the feedback I received on the previous version: [list feedback points]. Please update the design to address these points while keeping the existing flow and logic intact. Focus changes on [specific area] and avoid changing [elements you want to preserve].
Workflow at a Glance
| Stage | Tool | Purpose |
|---|---|---|
| Context and research | Gemini | Clarify ideas, gather inspiration, explore approaches |
| Prompt generation | Gemini | Create a detailed Figma Make prompt |
| First iteration | Figma Make | Interactive low-fi prototype, test flow and logic |
| Feedback and iteration | Figma Make | Refine based on review, align on direction |
| Stakeholder review | Figma | Share prototype and collect final feedback |
Closing Tips
A few things I learned from this process that will save you time:
Preparation is everything. The quality of your first iteration is directly tied to how well you briefed the AI before asking it to generate anything. Do not skip the context dump stage.
Stay low-fidelity on purpose. Resist the urge to push for visual polish too early. A working wireframe that communicates logic and flow is more valuable in early stages than a beautiful screen that does not move.
Let the AI suggest. Some of the best ideas in this project came from Gemini pushing back or offering alternatives I had not considered. Treat it like a collaborator, not just an executor.
Track your iterations. It is easy to lose track of what changed between versions. Keep a short note of what each iteration was trying to solve so you can move forward with intention.
Know the limits of your tools. Figma Make is great for speed and exploration. It is not great for pixel-perfect control. Use it for what it is good at and do not fight it.
Is this ready for development? Not yet. But in one day we went from a meeting to a working prototype that showed exactly what direction we wanted to take, what messaging was needed, and how the experience could flow. That is a huge win.
Use AI as an advantage. Learn how to research, how to build prompts, and how to express your thoughts clearly. It will elevate your work more than any single tool ever could.
Want to See This in Action?
I am thinking about putting together a video or a workshop where I walk through this entire process in real time, from the context dump all the way to the final prototype. If that sounds useful to you, let me know! Drop a comment below or send me a DM on Instagram. If there is enough interest I will make it happen.