AI support for building, testing, critiquing, iterating, and refining design work

On my last product design project I went all-in with my AI workflow, and I wanted to share it here. Maybe it's useful to someone, and I'd love to get some tips to improve it.

I went from nothing to a fully working dashboard (6 pages, realistic data, all features functional) in days. Including UX research throughout the entire process. That's 5x faster than before.

Here's what my AI workflow looked like:

Discovery research
Stakeholder interviews and a user interview on the current tool. I record and transcribe every conversation. Then I use my own Claude skill to clean up those transcripts, summarize them and pull out to-dos. That output goes into 2 documents: 1 cleaned-up transcript, which I can still search through later. And 1 with a summary + to-dos.

Design exploration
I made 3 design directions in Figma (partly in Claude Design), discussed them with the team and stakeholders, and picked 1.

Prototype & testing
I then built a working prototype, still based on a lot of assumptions. Showed it to users. Again transcribed every conversation and ran it through my skill. That output goes straight into prompts for Claude Design.

Iterating
adjusted the design based on what I heard. Showed it again, transcribed again, prompted again. Research and design stop happening one after the other and start happening seamlessly alongside each other.

Live testing
I used Netlify to publish the prototype on a free link. Users shared their screen during a call, and I could watch and ask questions. It had fake data they could actually edit themselves, and immediately see the effect on the dashboard. It was simply a fully working dashboard.

Handoff to dev
once the design was fully signed off, I built it out further with Claude Code and handed it over to the developer via Github.

Result
5x faster than half a year ago. Not because AI takes over the work, but because the distance between insight and iteration gets a lot smaller. And I spend way less time on write-ups, re-reading, and typing.

Curious if this sparks any ideas to add (or drop) from this workflow!
And if you want to use my transcript skill, I can send it!

How I use AI in my daily workflow to move faster!

The AI Step: I use Figma Make to generate a rough layout instantly. It saves me time on brainstorming and gives me a quick starting point. I just need to understand the project and also the direction the client wants!

The Human Step: I redesigned it to match the design style and what was wanted.

So AI does the most part, and I do the cleaning because I am a UI designer (I love to have clean interfaces).

What do you think of this workflow, and how do you design lately?

Plan mode

Plan mode is a new opt-in mode that helps you shape that direction before generation starts. Make takes a look at your project, asks a few clarifying questions, and drafts a plan you can edit, refine, and approve before anything gets built.

It's most useful for complex work—multi-section layouts, detailed specs, design imports—where getting aligned upfront leads to noticeably better results. For simpler prompts, you can skip it and build directly.

Turn it on via the dropdown in the prompt box or the /plan command. Because plan mode does extra work upfront, it uses more AI credits than a standard build—you'll see an estimate before you commit.

Web search & fetch

Make can now pull live context from the web mid-build. Search broadly or fetch a specific URL to ground builds in current content. Tool-call approvals let you review before anything enters your session.

Queued messages

Stack follow-up instructions while Make is still generating. Edit or delete them before they commit, and they'll send automatically once the current build finishes.

Okay, here’s something I didn’t expect…

AI isn’t reducing the need for designers.
It’s exposing the gap between them.

For a long time, being “good” at design meant:
- Clean UI
- Knowing Figma
- Following patterns

That was enough.
Now... AI can already do that.
And honestly… sometimes faster than we can.

But here’s where it gets interesting.
AI doesn’t know when it’s wrong, it gives you something that looks right, feels right, sounds right. Until it isn’t.

And that’s the moment where the real skill shows up. Not in how good you are at using tools, but in whether you can say:
“This is not good enough.”
“This misses the point.”
“This is what it should be instead.”

That takes:
- Judgment
- Taste
- Experience
- A point of view

Not just execution.

I’ve started noticing this in my own work.
The hardest part is no longer creating something.
It’s deciding what should be created.

So next time AI gives you a polished answer…
Don’t accept it immediately. Push back, rewrite it, make it better, then push back some more.

I noticed that the juiciest answers always come after 1-2 hours of discussion and rabbit-hole exploration with the models.

AI is a tool. Use it.
But design is still strategy, judgment, iteration, and craft.

I presented my custom AI design workflow to our design team two weeks ago. #ClaudeDesign launched 𝙩𝙬𝙤 𝙙𝙖𝙮𝙨 𝙡𝙖𝙩𝙚𝙧 and had a lot of overlap with what I built. I sat on this post for a while. 🧘‍♀️

Decided to share because building it taught me something about where custom workflows are still optimal (for now).

𝘢𝘭𝘪𝘦𝘯𝘦 𝘾𝘭𝘢𝘶𝘥𝘦 𝘿𝘦𝘴𝘪𝘨𝘯 𝘸𝘪𝘯𝘴
Seed of an idea → something “good enough” to move a conversation forward, fast. Built-in sketch tool (hey, that’s the “digital napkin” concept), consistent format, easy to share and comment on. For early ideation, it’s genuinely convenient.

𝘢𝘭𝘪𝘦𝘯𝘦 𝘘𝘶𝘴𝘵𝘰𝘮 𝘴𝘬𝘪𝘭𝘭𝘴 𝘸𝘪𝘯
𝘄𝘩𝘦𝘯 𝘪𝘵 𝘴𝘵𝘢𝘳𝘵𝘴 𝘣𝘦𝘧𝘰𝘳𝘦 𝘨𝘦𝘯𝘦𝘳𝘢𝘵𝘪𝘰𝘯. My workflow begins with /𝘣𝘳𝘪𝘦𝘧-𝘤𝘳𝘦𝘢𝘵𝘪𝘰𝘯, a live or async brainstorm between me and AI that builds a comprehensive project context before anything gets made. In async mode, AI generates a questionnaire based on the problem area, the team discusses and answers it in a working session, then we bring the answers back to Claude Code and AI picks up from there. That context carries forward into every skill downstream.

𝘾𝘰𝘮𝘱𝘰𝘯𝘦𝘯𝘵 𝘧𝘪𝘥𝘦𝘭𝘪𝘵𝘺. Claude Design rebuilds components as HTML files rather than pulling from the actual codebase I uploaded, which sometimes isn't 1-to-1 with the actual design system — missing states, missing components, approximate spacing. Custom skills can go deeper, connecting to your real design system components and anti-patterns, so what comes out actually matches your product styles and adheres to your design system better.

𝘏𝘩𝘦 𝘴𝘺𝘴𝘵𝘦𝘮 𝘭𝘦𝘢𝘳𝘯𝘴 — or at least, starts to. Every time I run /𝘥𝘪𝘳𝘦𝘤𝘵-𝘩𝘪𝘧𝘪 and flag what's off, the anti-patterns file updates. The goal is for team decisions and feedback to feed back in the same way, still figuring out that part. But even the early version compounds.

That’s the thinking behind this experiment: how my custom skills actually chain together across the double-diamond design process.
/𝘣𝘳𝘪𝘦𝘧-𝘤𝘳𝘦𝘢𝘵𝘪𝘰𝘯 → /𝘤𝘳𝘦𝘢𝘵𝘦-𝘥𝘦𝘤𝘬 → /𝘮𝘪𝘥-𝘵𝘰-𝘩𝘪𝘧𝘪 𝘰𝘳 /𝘥𝘪𝘳𝘦𝘤𝘵-𝘩𝘪𝘧𝘪 → /𝘪𝘵𝘦𝘳𝘢𝘵𝘦 → /𝘤𝘰𝘥𝘦-𝘵𝘰-𝘧𝘪𝘨𝘮𝘢 → /𝘢𝘥𝘥-𝘵𝘰-𝘴𝘺𝘴𝘵𝘦𝘮
Each skill's output is the next skill's input. Human checkpoint and decision making at every stage. Also, you don't have to use all the skills. Already have a brief? Start at exploration. Already have hi-fi? Jump to /𝘪𝘵𝘦𝘳𝘢𝘵𝘦.

Check the 3min video below for the highlights of my walkthrough. 👇

Claude Design is genuinely great, but a custom workflow can be tuned to your methodology, your design system, your rhythm in ways a general tool can't, at least not yet. Until it fully closes the gaps (and I believe it will), there's room for each of us to systematize how we work with AI, not just adopt it.

Still a lot to explore for future experiments: one-click wireframe ↔ hi-fi switching, converting prototyping code → production, tighter design system compliance, and more.

Curious what gaps you're working around with Claude Design or anything else. Leave your comments!

How I use AI in my Design Workflow

My design workflow has evolved over the last few months. I truly believe this is the best time to be a designer; the things we can now cook will be legendary.

I decided a while ago to not use AI to create designs at the agency.

We’ve tried it. Some parts of the process are cool.  But we’re creating human-to-device connection here.

And our designs come from our ideas which come from our conversations with you.

I’ll continue using AI as a tool for planning, parts of building and testing.  But when it comes to design, AI-use will be minimal if at all.

Design is not something I’m looking to delegate to AI right now.