In this round‑table discussion, designers compare the collaborative strengths of Figma’s open canvas with the isolated nature of current coding tools, noting that design work benefits from real‑time, shared exploration while code environments lack that fluid teamwork. They describe a workflow gap where mockups are built in Figma, then handed off to developers to recreate as production‑ready components, leading to duplicated effort across platforms. The group recounts experimental attempts to bridge this gap using AI: generating high‑level app maps with Claude, converting those maps into screens via Figma Make, and iterating on prompts to improve fidelity. They emphasize that effective AI prompting mirrors giving clear direction to a human teammate—specific, detailed instructions are essential, yet even precise prompts still require post‑generation editing because the tools are not yet mature. The conversation highlights micro‑automation opportunities, such as auto‑generating hover states, tags, and other repetitive UI details, freeing designers to focus on creative decisions. One participant frames AI as a reflective partner that can ask probing questions, helping designers break out of habitual patterns. Overall, the participants are cautiously optimistic: they see AI’s potential to reduce manual translation, accelerate creative work, and differentiate products, while acknowledging current limitations and the need for better integration between design and code.
Chapters
With Figma, the collaboration's really effective where people can go into the canvas together, they can leave comments, they can talk, they can chat. But then like with these coding tools, you're like in your own little project. Unless I'm unaware, it's not that other people can jump in and like collaborate in the same way. And I think show examples on the side, like, hey, well, what if it was like this? Like, it's hard. It's hard to, it's like, well, no, I gotta do my own. Cloud Code thing over here, and then we just kind of compare them side by side or whatever.
Yeah. Right.
Like the process, that kind of process hasn't really been figured out.
No. And I think it's still two different, it's two different like experiences. In Figma with the open canvas, a lot of that is creative exploration. Like, and I still think that's the main intent of Figma is like, I'm, I'm trying to figure things out. I need a, I need everything to be unlocked. I'm not going to engage with auto layout yet. I'm just trying to see what does this look like? What does it feel like? How could these things? Be built together. And then once those decisions have been made and you're feeling comfortable, now you get into this more rigidity that you're trying to establish because now you're trying to ship to a dev to build these components out. And so you're using variables, you're using styles, and you're building out different states for things. But then you kind of hand it off to dev to translate that into the, the real element. Like we're still building a simulated element and then they're building the real element. And I think the refinement to building the real element, that gap closing is going to be really nice because then we can help out with that and we can help drive that instead of essentially building the same thing 2 times because that's what we're doing now, or 3 times or 4 times depending on how many platforms you're working on.
The app mapping process we do, is there currently a way to kind of like connect the dots there, mock up something with this app map data? Have you done something like that before?
We've attempted to, it has not been great.
Not been great.
What did you try?
I'm just kind of curious, maybe share your experience with that.
We've taken sort of some of our app map information and I mean, this was a rough proof of concept, so put some parameters on it. It was a rough proof of concept. We didn't go too deep on it, but we did it two different ways. We were one, hey, we have this brief, this is the information we have. Here are some documents, create an app map for us. It gave us kind of a high-level app map or a flow diagram, but it missed a lot of the nuance and the way that we were working. And it— I want to look into like Lucidcharts and see if there's a more— something that I can craft more prompts to get it to what I want it to be. The— when we were doing Claude to Figma, it was a one and done. It shot out the app map and they were done, and you have to kind of reprompt it and do that. Sure. Taking the app map item and having it spit out a screen. When we did that in Figma Make, it, I mean, it was okay. It still created a bunch of gaps for various things. It's a lot better when you can give it the app map information and then like a mockup.
Oh, okay.
Like some of the explorations that you had done, Brianna, where we actually gave it the mockup and then it was a lot more informed about what it should be, that worked a lot better. But it's still like you spend a lot of time prompting afterwards to, to really craft, to get it into the spot that you need it to be.
Yeah, it's like wrestling it into where you want it to go and you could have just made it yourself.
Well, yeah, I think like the way I've thought about using AI and like prompting it is you have to, having had experience leading teams and giving direction, you have to give it direction as you would give a team member direction. And you have to be very specific if there's things that need to happen. So you can't leave out details because in the sense of like working with a person, they're just going to fill those in. And what you have in your head and what the person you're working with or the agent you're working with has is not going to be necessarily the same. So being very specific about what you want in that prompt is going to be helpful. But then also I think to Billy's point, the tools aren't quite there yet. And so what ends up happening is even if you're that specific, your output still has to be edited. And then it's sort of in the case of like if you're working with an intern with no industry experience at all and they just don't know what they're doing. So I think there's like this space of the AI has to improve and the more you use it, I think you get better at prompting, but there's still just like this gap in the tools. Like it, it can't do certain things yet. That doesn't mean that it won't be able to in the future. And I think that's exciting too, because I think as designers, creativity is important. And most of us want to hang out in the creative space and not in like this tedious, like data heavy or like number crunchy space that sometimes you have to do in order to build features and make sure things are correct? So the more opportunity we have to be in that like free flowing, open canvas in Figma, like I think the happier folks are. So allowing AI to do some of those tasks and help us get there quicker, I think will allow for hopefully more creativity, which I think will help solve for some of the like AI sameness that we're seeing now, where it's like a lot of the stuff that is being output from AI in that creative space is very like same, same. And having uniqueness in your product and your brand is gonna help you stand out. So you don't want everything to look the same, cause then there's no differentiator.
There's like all these micro opportunities for automation that we can leverage AI for. That's really exciting. Like we always joke, oh, okay, you've Designed out what this brand and what this look should be. Now you have to go in and make the hover state for all of these and then the tags and all this other stuff. And like getting a first draft of all of that automated and across the line is a very exciting prospect. And then, you know, being able to hone that afterwards, cuz a lot of that stuff isn't a lot of fun. It's not terribly exciting to do. It's a lot of manually translating things. So the less we have to manually translate stuff, I think that'll be, that is gonna be more, uh, where we can then be more expressive and try different things and be a little bit more interesting with our buttons and inputs and, and brand decisions. Because we don't have to worry about like, oh, is this gonna fit in the box that we have? Because, well, the tools will help us get it in there.
Yeah, it's almost like the details are really where the impact's made, as long as you have the intent behind what you're doing and understand your craft really well, like those tools will just help you do your job better as long as you're, you're, you're following like the basic principles that actually matter.
A lot of times we're looking at this to solve and automate things that we've done, but also using it to ask more questions of yourself when you're doing different design work. That's something that I've recently been doing a little bit more of is is using Claude as sort of like, hey, if I was going— if you were going to go do this, what are some questions that you would ask me to give you the correct answer for this? And then using that as a prompt for me then to think through, oh, okay, well, maybe I should add this, or maybe I should consider this, or maybe I should do something like this. And that's been really useful as like kind of a sounding board almost instead of— or just to expand your thinking a little bit because sometimes we get trapped into our repeatable patterns that we've done, and it's always good to kind of push ourselves a little bit.
