← Return to Index Archived April 22, 2026
The Lead — Apr 22
HOW I AI · CLAIRE VO

What Claude Design is actually good for (and why Figma isn’t dead, yet)

27m / April 22, 2026 /aiproducttechnology / Transcript sourced from openai
All episodes from How I AI →·Podcast website →·Listen on Apple Podcasts →

Overview

In this mini-episode of How I AI, Claire Vale reviews a wave of new AI design tools, focusing primarily on Anthropic’s Claude Design and OpenAI’s new GPT Image 2 model. Her core question is practical rather than hype-driven: which of these tools are actually useful for real product, marketing, and brand workflows, and where might they start to replace parts of the traditional design stack.

Claire’s conclusion is nuanced. She does not argue that Claude Design will fully replace Figma, but she does believe it could replace parts of the pre-engineering prototyping workflow, especially where branding, design-system consistency, slides, and marketing assets matter most. She also sees GPT Image 2 as a meaningful step forward for brand kits, layout-heavy image generation, and text rendering.

Key Takeaways

A major insight from the episode is that design systems are becoming the foundation for AI-native design workflows. Claire highlights Claude Design’s ability to import a design system from HTML, images, and optionally GitHub, then convert that into a structured format of typography, colors, UI components, and brand assets. This matters because one of the biggest weaknesses of AI prototyping tools has been their inability to reliably match an existing brand or product interface.

She also points to an emerging industry shift toward standardized, machine-readable design documentation, referencing Google Labs’ new design.md standard. The implication is that design systems may soon function for design tools the way structured documentation already does for code assistants.

Claire is particularly impressed by Claude Design’s fit for marketing use cases. In her tests, it performed well when generating high-fidelity landing pages and especially slide decks that adhered to a brand system. The slide workflow stood out because it combined source content with design rules to produce polished presentations, suggesting a strong use case for product marketing, enablement, and customer-facing materials.

Another notable takeaway is that variation generation is a powerful product feature. Rather than forcing users to over-specify prompts, Claude Design can propose multiple directions up front. Claire frames this as especially useful for non-designers who know what they like when they see it, but may struggle to articulate detailed revisions.

On GPT Image 2, her strongest praise is for layout and typography, two areas where image models have historically struggled. She found it notably better at producing branded visual compositions, multi-panel brand kits, and infographic-style outputs with readable text. While not a replacement for professional designers, it appears to be a strong starting point for teams that need fast creative exploration.

Practical Steps

  • Create a structured design system for AI tools. Break your brand into typography, color, UI components, logos, and other core assets. Even if you do not use Claude Design, this structure will likely improve outputs across AI design tools.
  • Test Claude Design on marketing pages first. If your team has struggled to get AI prototyping tools to reflect your brand, start with a landing page or campaign page rather than a core app workflow.
  • Use content-to-slides workflows for enablement. Take an article, PDF, or internal doc and pair it with your design system to quickly generate branded training decks, customer presentations, or sales materials.
  • Ask for multiple design variations early. Instead of endlessly refining one prototype, generate three to five options and choose the direction that feels strongest.
  • Use GPT Image 2 for early-stage brand exploration. Prompt it to create moodboards, brand kits, or layout concepts, then refine with reference imagery to push the outputs closer to your desired visual identity.
  • Experiment with playful references. Claire shows that when you remove rigid brand constraints and provide a strong stylistic prompt, Claude Design can produce highly creative concept work and surprisingly strong copy.

Notable Quotes

  • Claire Vale: “I think it might actually replace some of these prototyping tools that everybody’s using before they go to engineering.”
  • Claire Vale: “What I think makes Claude Design really interesting is they’re making the design system a first-class citizen of this design tool.”
  • Claire Vale: “It writes exceptional copy… which is such an underappreciated part of building a great prototype or design.”

Full Transcript

Source: openai 27m runtime

Welcome back to How I AI. I'm Claire Vale, product leader and AI obsessive here on a mission to help you build better with these new tools. This week, there has been so much released in the AI for design space. Whether you're looking at Claude Design trying to decide if it's really going to replace Figma or want to know if the new GPT image 2 model is actually that good, I've got you covered with a mini app where I'm going to show you what I really think about these new design tools and how they'll work for your business and a couple fun use cases. Let's get to it. This episode is brought to you by WorkOS. AI has already changed how we work. Tools are helping teams write better code, analyze customer data, and even handle support tickets automatically. But there's a catch. These tools only work well when they have deep access to company systems. Your co-pilot needs to see your entire code base. Your chatbot needs to search across internal docs. And for enterprise buyers, that raises serious security concerns. That's why these apps face intense IT scrutiny from day one. To pass, they need secure authentication, access controls, audit logs, the whole suite of enterprise features. Building all that from scratch? It's a massive lift. That's where WorkOS comes in. WorkOS gives you drop-in APIs for enterprise features so your app can become enterprise ready and scale up market faster. Think of it like Stripe for enterprise features. OpenAI, Perplexity, and Cursor are already using WorkOS to move faster and meet enterprise demands. Join them and hundreds of other industry leaders at WorkOS.com. Start building today. I have to start today's mini-episode with a little bit on Claude Design. In case you missed it, last week Anthropic released Claude Design and it's their web-based design tool. Really focused on a couple areas. Prototypes, including wireframes and high-fidelity prototypes. People are asking, is this going to replace Figma in the design flow? Spoiler alert, I think it might actually replace some of these prototyping tools that everybody's using before they go to engineering. And also is taking a swing at slides and videos, which I think is kind of interesting. And also has templates for animations and videos, which we can also show a little bit of on this episode. But what I think makes Claude Design really interesting is they're making the design system a first-class citizen of this design tool. So the first thing I did was actually go in and just import my design system. And I think this is really smart because the number one complaint I know people have about using prototyping in the product development lifecycle is it never matches my brand. It never matches our design system. And these prototyping tools, while I think great, you know, the Lovables, the V0s, the Bolts, etc., also have this concept of design system import. I really haven't found they get perfect at replicating my existing design in my codebase. And so I really wanted to test this use case for Claude Design and see if it actually works. So the first thing I did was really import my design system. Importing a design system is pretty easy. And I'm actually going to show you how to do it by importing Lenny's newsletter design system. So if you see here, this is what I filled out to import Lenny's design system. Sorry, Pal, we'll see how this looks. I really just went to his homepage. I clicked save on the HTML. I saved the logo. I saved our little campfire here. I saved this nice little graphic and I just put it in here in the Claude code create your design system form. So I said, it's Lenny's newsletter. I don't have access to his GitHub. If you want to give me access, friend, I can, I can make this even better. I uploaded just a saved HTML and some images, and then added those fonts, logos, and assets. I didn't even put in additional notes here. I just left it blank. And then as you can see, Claude Design goes ahead and starts exploring the provided materials and wants to understand the newsletter and then build out your design system in the format that they need. And what's really interesting here is they warned you. They say, this is going to take about five minutes. And it did. It took about five minutes. And you look along the side on the reasoning. It's pretty interesting to see how it extracts the design system. It's looking at the HTML and key images. It has a visual picture now. It has a sense of what core colors and other things it's going to work. And then it's going to build out this structure of a design system, sort of in a skills and documentation mode that then Claude Design is going to use. And look at here. I think it looks pretty good. If you look at this, it looks pretty comparable to this. I would say the only difference is Lenny uses sans-serif fonts for his headers. So what I might say is, you know, we typically use sans-serif fonts for the main headers other than top level. Switch that. But otherwise, it looks good. The other thing that is pretty cool here is there's, you know, the newsletter kit, the typography, colors. So you can actually see how it breaks down the idea of a design system. Even if you're not going to use Claude Design, the idea of breaking your design system for any AI tool into these components, UI kits, typography, color, components, and brand marks is really interesting. In fact, just today, Google Labs released the design.md standard, which they're trying to make sort of a standard Asian.md or skills.md standard across how you would describe a design system for use in Google's design tools but also other AI design tools. And so you're really seeing a lot of these solutions move to the concept of a structured design system that then can be used by AI to render consistent high-quality UI against your experience. So the design system is really interesting. I also created one from ChatPRD. The one that I did here, I used the actual GitHub. So I hooked up the very specific GitHub directly to this app. So it pulled in some pretty specific components and UI across our app and our marketing site. So you can put as much or as little as you wanted to these design systems, but I think it's a really interesting flow. And then going back to the core capabilities of Claude Design, I just want to show you a couple things that I thought was really interesting. One is, I think, its ability to adhere to a design system for marketing-specific landing pages and websites is really good. So if you were on the marketing side and you had been sitting out doing prototyping because it's so hard to get your brand assets into a really nice workflow, I think Claude Design is a really great one for you because it's particularly good at that. Okay, so I'm going to use Lenny's design system and really show you what this looks like in practice. So let's say Lenny is going to come into competition with ChatPRD and we're going to create the Lenny PRD builder using the Lenny design system. And we're going to have a head-to-head compete, Lenny or Claire, who writes better PRDs. Well, I'm going to use Claude Design, the Lenny's newsletter design system, and I'm going to say PRD maker. And I'm going to click create on this high-fidelity prototype. I'm going to start with the design system. You can see it's already selected down here. It's an interactive prototype and we want to do hi-fi design. And I'm just going to use my monologue to write in here what I want. Please help me make a landing page for a PRD generator and AI PM coach powered by all the data in Lenny's newsletter adhering to the Lenny newsletter design system. Make it awesome. Make it awesome is the design version of please Claude make no mistakes. So I'm going to go ahead and set that in. And if you're looking for a little bit of fun, we have a couple new loading messages here over in the Claude AI design loader. It's not flipper jittering or whatever it used to be in Claude code. It does things like saute onions. Very funny. Okay, so it's going to ask me about the purity generator landing page. Again, I really love this Q&A functionality that's in Claude code and they've made it very nice in Claude Design. The product is called LennyDoc and it's going to help you an AIPM in your pocket. It's trained on Lenny's archive. Lenny, please don't compete with me. I'm tired. All right. Who's the primary audience? New and aspiring PMs. It's going to be pure Lenny. Decide for me on what sections I should include, how interactive should it be. I'm going to do small live interactions, but not the full thing. Again, this is for a prototype. I'm going to let it decide pricing. I love this. If you haven't seen this in Claude Design, they give you little wireframe hero style directions and I even got one of these that said, tell me what appeals to you. I love that so much. And for me, I'm just going to go with this. This appeals to me. I don't know. I just like it. And then I'm going to just let it decide. And then here's another really cool function of Claude Design. It gives you variations, which as somebody who has spent so much time in A-B testing in her career, I think is so fun. You can decide how many options you want. You can get one option, you can get three options, you can get five options. It defaults to three. And I think this is really smart from a design tool perspective because often those cycles of, no, make it better or make it different can be very slow and most people come into Claude Design probably don't have the ability to articulate exactly the changes they make. thing is Claude Design can make slides. So I chose that landing page, Lenny's newsletter design system. And I actually dropped my OpenClaw article that I wrote for the newsletter in as a PDF. And I said, let's build a deck teaching people how to set up OpenClaw based on this article. And this is where I think it really shines. It looked at the design system, looked at the content and built a really nice presentation for me to present OpenClaw and how to set it up. And you can see here, it just looks really nice. And so if you're someone that has to put together decks for customers, you have to put together instructional material, and you really want that to match your branding. I think product marketers, you really need to pay attention for things like training, enablement, and customer decks. I really think this does a very nice and elegant job of designing a deck. And even look at this. I had a command in the terminal as part of my article and it actually put in a really cute fake terminal here with a little blinker. And this, the fun parts of the Claude Design slides is they're actually just code behind the scenes. I think this might be where we're going with slides overall is slides just end up being code like everything else. And so you can get these cute little design elements that I think are really delightful and fun in the app. So if you're looking for a really practical use case where I think it does quite well, take straight content and turn it into a beautiful deck matching your brand. And then my final, very important Claude Design use case, it gets very fun to imagine crazy versions of your web experience or website. When you give Claude Design no design system, it does the best. So I asked Claude to make a 90s Geocity style version of the Lenny's newsletter homepage. It did that. It is called Lenny's Product Zone and it is pretty incredible. And even the tweaks component that we had that was so nice in the polished landing page version is now real ugly. I'm going with bricks. I'm going with Comic Sans, of course, and I want all the things. So I'm going to leave all these tweaks here and look at this beauty. Your OKRs are cringe and seven ways to fix them before Q3. And this is a secret power of Claude Design. It writes exceptional copy. So if you let it run on copy, which is such an underappreciated part of building a great prototype or design, it's going to do a really, really good job. And so I highly recommend finding a reference style. Jamie Gannon makes a really good point on this in her episode of How AI Is if you understand reference styles, you can reference and explain what you want, but let it go pretty wild without a design system. And then you will get very, very cool designs. So I'm going to, as soon as Lenny gives me access to his GitHub, I'm going to push this right up. I think it's going to convert, convert like crazy. So let's ask him to ship it right in the comments if you think this is what Lenny should redesign his website to. So those are my three Claude Design use cases. Import your design system and create a new, more marketing style landing page. Use content plus your design system to create really, really beautiful slides and then go ham on an ugly redesign of your website or ugly or beautiful, depending on what you think. And fax us at 1-800-PRODUCT, which is exactly what this landing page says is the way to get in touch with my friend Lenny if you think he should shoot, ship Lenny's Product Zone. Okay, I just want to show you two more things really quickly before we go. And those things are how I think you can use the new GPT Image model 2 to get some brand work and also some fun image work done. Okay, so just today ChatGPT OpenAI released ChatGPT Images 2.0 model, which is new era of image generation. It is the first model that can do thinking and they've really focused on text and the ability to render text and objects accurately. And I just wanted to show you two use cases that I think are pretty interesting and how well I think the new image model from OpenAI performs. The first one is kind of back to what we were talking about before, which is brand. And I saw this prompt on X and I copied it for ChatPRD to see what we would get. And it's basically the idea of creating one of those multi-page brand kits for your company. So I asked it to create it for my company, ChatPRD. I just gave some general kind of brand design guidance off this. I had ChatGPT write this for me and asked it to generate it. And what I will say is it did a really lovely job in terms of text. It doesn't have that classic like GPT image text looks to it. So typography has gotten a little bit better. I think the, it's kind of amazing. It can do this nine grid layout of different parts of the brand kit. And I don't think it's bad. It's just not me. And so I loaded up some images from Mid Journey here that I think we use more in our brand kit. And I say, that's not really us. Here are some reference images. Update the brand kit. And let's see what it does with that. I'm going to make it create an image. Hopefully, I'm hoping it should read these images, which are much brighter, much pinker, have this like sort of pixelated feel to them and create a brand kit that looks more like what I would want to use. And I think this is just such a neat workflow. I've heard a lot from people that they're really dissatisfied with the current image models for marketing purposes because it requires so much bringing together of images and text and typography and voice. And while I think something like a Claude design obviously can do that well at the web asset level, sometimes you want it at the brand or image asset level. And I think it's really interesting to see if GPT Image 2 can do this. So let's see what it generates. It's finishing up right now. Let's see how pink it is and if we love it. Oh my gosh, I actually really love it. It makes me so happy. As you can see, it took in this like sort of pixelated landscape-y images that we've been using for ChatPRD. It is certainly more pink and it has given me kind of a new idea of what we could do with this brand kit. Now, we got an amazing brand kit done by designers. I think it's probably scientifically 5 to 50 times better than this. But if you were looking for somewhere to start, I think that GPT Image 2 is quite good at this. And this doesn't have that very obvious AI image text tell on the typography. And I think something like this is a really great starting point for folks that are trying to get images done. And then finally, one last fun use case. It just finished over here, which is taking my image and doing my color analysis for the boys in the room. This is something that the ladies often do to figure out what colors look best on what they're wearing. And so I uploaded a photo and asked GPT Image 1, Image 2 to make me a color analysis. And it gave me warm neutral. My best colors are these earthy warm colors and even put me in those colors here, down here. Now, my face got real weird towards the bottom. My teeth are looking kind of funk over here. But overall, I would say this is pretty impressive. This is something that people spend a lot of time putting together for folks. Now, what I would say GPT 2 is, I am actually a dark winter. So I might come in here and say, no, I'm a dark winter. You are just confused by my old highlights. My hair looks much blonder in the picture than it naturally is. It's actually very dark. And so let's see if it revises to my exact color palette. But again, think about the combination of text, color, image analysis, layout quite lovely. And I haven't seen a model do this kind of layout work effectively. You know, I think Nano Banana had gotten close with infographics and things like this, but we might actually switch some of our infographics and graphic generation over to GPT Image 2 because I think it just looks a little bit more expensive and a little nicer. All right, here we go. This actually looks really great. This is the correct color analysis for me. My natural hair color is actually the one in the middle. Don't tell anybody. And these are the colors. Now again, my face gets really weird down here and I would never wear these clothes. I am not a business lady. I am a hacker podcaster. But in terms of the colors, in terms of the layout, in terms of the photography, the typography, this is pretty impressive and I would pay a lot of money for somebody to do this for me. So if you're looking for a fun Mother's Day gift out there, take a photo of your lady, one where she's looking good, and throw it in here and ask for her color analysis and then buy that person some jewelry. All right, so this is the design-focused spring 2026 drop. We looked at Claude Design and how good it is at importing design systems, creating marketing style landing pages and slide decks, and then translating your current website into something a little bit more fun. We looked at the new GPT model, GPT Image 2. Two things I think it's really, really good at. Layout and typography. It has really nailed those things and has also just up-leveled the quality of design because this is the first image model that really does some thinking. I think it's great. And we talked about Google's design.md standard, which they're trying to