Claude Design in Practice: Our Honest Take After 5 Days

Claude Design went live on April 17, 2026. We dropped what we were doing two days later and have spent the last five days using it. Not in a sandbox. Not as a research side project. In our real work, on real briefs, as a working web design agency.

Call this a field note rather than a tutorial or polished case study. Five designers, five days, a lot of strong coffee, and a tool that kept surprising us. We build landing pages, corporate sites, and web portals for a living, so we tried to break it the way clients break us.

Five days is nothing. I know. But sometimes five days is enough to tell which way the wind is blowing.

Claude Design is Anthropic's interface design system, built on top of the Claude model family. It is not an image generator. That one sentence matters more than most of the launch coverage gave it credit for, and it's the reason this thing lands differently than anything we'd tried before.

Midjourney and Adobe Firefly give you pictures. Claude Design gives you structure: a layout, a hierarchy, real spacing, components that know about each other, interaction logic that follows from the business context you fed in. You don't end up with a screenshot of a website. You end up with a design you can actually build from.

The practical consequence is that the translation step goes away. You're not looking at a beautiful reference image and asking yourself how to turn it into production UI. You're already inside the UI. That collapses the iteration loop in a way I wasn't prepared for on day one.

Claude Design vs Midjourney

Midjourney makes gorgeous images. For brand moodboards and early creative direction, it's still excellent and we still use it. But for interface design it was always a translation problem. Somebody had to sit down and rebuild the pretty picture as real components. Claude Design doesn't take over that use case. It takes over something further down the pipeline, where we used to spend most of our hours.

Claude Design vs Adobe Firefly

Firefly lives inside Creative Cloud and generates visual assets: product shots, background variations, compositional experiments. Web interface design isn't really what it's for, and I suspect Anthropic knows it. Comparing the two is almost a category error, but people keep asking, so here we are.

Claude Design vs Figma AI

Figma AI refines and extends what's already in your Figma file. It's a smart assist layer on top of manual work. Claude Design operates earlier. Before there's a file. Before there's a structure. At the point where you're still figuring out what the thing should fundamentally be. One works upstream, one works downstream, and most of the online debate about which is better seems to miss that they aren't really competing for the same minute of your day.

We didn't pace ourselves. From the first hour after it went public we were throwing things at it. Landing pages across industries we know well (fintech, logistics, healthcare), UI component systems, navigation architectures, design system foundations, brand identity directions. Generate, evaluate, push back, regenerate. Rinse and repeat until somebody got hungry.

What we tested

Landing pages were where we started, because a landing page is the sharpest test of whether a tool actually thinks. A good landing page isn't pretty by accident. It's built around a specific sequence of things the visitor needs to believe before they'll act. Claude Design doesn't just spit out something that looks like a landing page. It reasons about that sequence. The first time we saw it structure a cold-traffic layout differently from a warm-traffic one, using the same product brief, without being told to, the room went quiet for a second. That was the first real pause.

UI components were next. Component libraries, data viz sets, form systems, nav patterns. The consistency within a single session was better than we expected. Things felt like they came from the same system, not assembled from a bunch of unrelated Pinterest boards.

Day five compared to day one

On day one we were poking at it. Finding edges. Trying to see what it couldn't do.

By day five we were directing it. We'd picked up the grammar of working with it: how to write a brief so the first pass is already close, how to push back without confusing it, how to use constraints as acceleration instead of guardrails. The gap between what we fed in and what came back narrowed noticeably across those five days. That's usually a sign of a tool with real depth underneath, not just a clever demo.

We've used a lot of AI tools in real workflows. Most of them speed up one specific task and leave the thinking part alone. Claude Design did something different, and I'm still chewing on what to call it.

Structural reasoning, not just pretty pictures

Claude Design doesn't just build what you ask for. It reasons about why the layout should be structured the way it is. Ask for a SaaS landing page aimed at an audience that's never heard of the product category, and it puts education before social proof. Ask for the same product aimed at people already in the category, and the structure moves around. This is not a visual tool with a chat box on the side. It's a design-thinking tool that happens to produce visuals.

The pushback is real

More than once during the week, Claude Design flagged that our brief was ambiguous or contradicted itself, before it generated anything. Not by failing. By asking. That's closer to working with a senior designer than a production tool, and I'll admit it's not always welcome when you're in a hurry. The results were better every time we slowed down and answered properly. Annoying. Also correct.

Iteration speed

Direction changes that would normally eat half a week happened in minutes. Not because the tool does the same operations faster. Because the operation itself is different. Changing a layout direction isn't moving elements around. It's rewriting the brief and letting a new layout fall out of it.

Context stays put

Inside a session, Claude Design remembers. Components built in the same session share typography, spacing, interaction patterns. That sounds minor if you haven't done this work for a living. Consistency across a design system is one of the most labor-intensive parts of what we do, and having it handled without babysitting is a new feeling.

Five days in and the main thing I can say with confidence is that we haven't seen the ceiling.

We haven't stress-tested it on the hardest work we do: multi-stakeholder discovery processes, large-scale portal architectures, brand systems that need to hold together across 12 different touchpoints. Everything we've done so far is still exploratory by the standards of a real client engagement.

And even at that shallow end, it has already changed how we think about the opening phase of every project we'll touch from now on.

The tools that turn out to matter don't always arrive with trumpets. Sometimes they show up quietly and a few weeks later you notice you've stopped doing something you used to spend hours on. I think Claude Design is one of those, and I think we're at the very start of figuring out what that actually means day to day.

So we're integrating it. We're building our process around it. We're treating it less as a novelty to evaluate and more as a capability we have to develop. That shift happened in five days. If you haven't tried it, Anthropic's official getting-started guide is the fastest way to form your own opinion.

Let me say this up front. We actually like Figma. Figma changed collaborative design in a way that was different in kind, not a small upgrade. Before Figma, design collaboration was a pain. After Figma, it mostly wasn't. That's a real contribution and I won't pretend otherwise.

But Figma has an underlying problem that Claude Design makes obvious, and I don't think enough people are saying it out loud.

Everything inside Figma still needs a human to make every visual decision, move every element, define every component relationship. Figma AI helps, but it's help bolted onto a fundamentally manual setup. The design file is still the center of gravity. Your hand is still on every pixel.

Claude Design doesn't work that way. It reasons about structure before generating it. The mental model is inverted. You define the goal and the constraints. The structure follows. That's not a feature difference so much as a whole different way of working, and I think that's where the discomfort comes from when designers try it for the first time.

Adobe's problem is harder

Adobe's position is tougher than Figma's at the level of how the business is built. Creative Cloud is enormous: Photoshop, Illustrator, After Effects, XD, InDesign, Premiere, Lightroom. The installed base is massive, the enterprise contracts run deep, and whole decades of workflows sit on top of it. Firefly really is impressive at what it does.

The thing is, Adobe optimizes for what Adobe already is. Firefly generates imagery. It does not reason about interface structure. The whole suite was built for a world where creation meant manipulating visual elements with skilled hands. In that world, Adobe has no real competition.

The question I keep coming back to is whether that's still the world we're in.

AI-native tools, the ones built from scratch around conversational design generation instead of visual element manipulation, don't need to beat Adobe on Adobe's terms. They just have to make Adobe's terms feel a little dated. And that's a much easier game to win.

In 2007, Nokia built around 40% of all the mobile phones sold on the planet. They had the supply chain, the distribution, the brand, the carrier deals. They had every advantage you could want in the mobile phone business. Then the iPhone showed up and redefined what a phone even was, and Nokia's advantages turned out to be advantages at building the old thing.

BlackBerry had a different kind of lock on the corporate market. Secure mobile email, they basically invented that category. Enterprise IT departments standardized on BlackBerry. The loyalty wasn't just habit, it was baked into how enterprise IT worked. By 2013 they were negotiating for survival.

Nokia and BlackBerry weren't badly run, and that's the whole point. RIM's engineers were excellent. Nokia's manufacturing and distribution really were world-class. What took them down was that dominance in one paradigm is no defense against a paradigm shift. When the ground moves, your moat becomes the thing slowing your pivot.

What this has to do with design tools

Figma's moat is that every designer already knows it. Every design team has files, components, libraries, and processes built inside it. Every developer handoff workflow connects to it. That moat is real and it is deep.

It's only a moat if the new way of working still requires Figma. If the design process actually shifts, from "designer manipulates elements in a file" to "designer directs an AI to generate structures from intent," then Figma's moat sits in the wrong place. Expertise in the old model does not automatically transfer to the new one, and that's the part that usually gets glossed over in these conversations.

I'm not predicting Figma will fail. I'm saying Figma faces the same challenge every dominant tool faces when the ground moves underneath it. The companies that came through moments like this in one piece did it by treating the new approach as a first-class priority, not as a feature to bolt on. The ones that didn't, we remember their names for different reasons now.

No. But "replace" is still the wrong frame, and the way the question is usually asked tells you more about the asker than about the job.

Claude Design compresses execution. The part of design work that was always more craft than thought, making the first layout pass, iterating on designs, producing component variations, is now a lot faster. What's left, and what gets more valuable, is judgment. Is this structure right for this specific business problem? Is the brief wrong and should we push back before we execute? What are the stakeholder dynamics that the brief never captured?

The designers doing well right now are the ones treating Claude Design as a real lever, not a threat. They're writing better briefs. They review outputs faster. Most of their energy goes into the decisions that actually move the needle on whether a website works: conversion logic, information hierarchy, how real humans behave when they land on a page.

The designers who will struggle are the ones who built their professional identity around production velocity instead of judgment quality, not the ones who lack technical chops. And I say that with sympathy, because the industry told them for years that production velocity was the thing.

This is the same story as every previous wave of tool change in our business. Photoshop didn't replace designers. It replaced people whose job was execution without judgment. Designers who learned Photoshop got more powerful. The ones who didn't adapt woke up one morning and the field had moved.

Same thing is happening now, just on a much shorter clock. Five days in, we're already working differently than we did a week ago.

We started working with Claude Design on April 19, 2026, two days after it went public. Five days later, this isn't a review of something we evaluated. It's a note about something we're integrating.

We're impressed. Not in the hedged, sensible, cross-your-arms way that professional skepticism usually produces, but in the straightforward way that comes from watching a tool do something you didn't think was possible yet. The structural reasoning, the iteration speed, the way components stay consistent inside a session: that isn't polish. It's a different way of thinking about what the design process even is.

Figma will respond. Adobe will respond. Both have the money and the talent to build compelling AI-native capabilities. The open question is whether they can move fast enough, and far enough from their existing ways of working, to pull it off. Honestly, I don't know the answer, and I don't think anyone does yet.

Nokia. BlackBerry. Those names still mean something, and so does how fast things moved once they started moving.

As an agency, we work with AI as our core operating model, not a side capability. Claude Design is the most significant tool release we've seen since Figma itself. We saw it on day one. And from here, we're building on it.