Config 2024: The future of design to code in a world of AI (Kris Rasmussen & Meredith Black, Figma)

It's me again.

Hi. Hi, Kris.

So glad you all are here. I bet most of you have a VP of engineering, a CTO, somebody to do the development work, likely. You're also probably curious about where generative AI is going and what might happen next. And you're probably curious about where Figma will go next with all of it. So I'm really excited for this next conversation with Kris Rasmussen and Meredith Black from Figma. Please join me in welcoming them to the stage. [FOREIGN] Alexander.

Hi everybody, hope you had a good lunch. All right, Kris, I am really excited to have you up on stage to talk about the connection between design and code. I think there are probably a lot of us in the room right now who are design leaders, product leaders, and leaders who have worked closely together on building products. And I think this connection of dreaming of what we want to build and actually building it is something that's merging closer and closer together. So I'm new here at Figma, just got here 3 months ago. Last year I was sitting in the audience and we had just come out with Dev Mode. I'm curious, why did we build Dev Mode?

Yeah, I, I think externally we talk about this story about our weekly active users, which is a very true story. So when we look at like our weekly active users inside of Figma, Um, only 30% of them are actually product designers. The other 70% of the people who use Figma every day are people interacting with product designers in order to bring those ideas to life and bring them to production. And, um, as a result, we knew that we needed to create this space inside of the product where we could cater to the needs of another third of those weekly active users, the actual developers who are looking at those designs and, you know, giving feedback on those designs and bringing those designs to life. But the actual story starts like way back when we first launched the product. I mean, the whole reason we thought there was an opportunity here is because we felt like the old way that designers and engineers worked was just fundamentally broken. It was this slow, laborious process of designing something and then handing it off and rarely kind of going back and forth. And it just felt like there was this huge opportunity to just dramatically improve the collaboration by building a design tool on the web and doing all the hard work it takes to actually build performance graphics application on the web. And so the reality is this is— this has been in our sights for a very long time, but it wasn't until we really felt like the product design tool was in a really solid place that we thought it was worth really kind of going deep into the needs of developers.

Got it. Yeah. I think also what's really interesting, just nerding out from a DesignOps perspective, like one of the things I was really excited about is when this happened to somebody in my role, like we own the budgets and like we own like tooling budgets and like we do all this stuff, right? And so for me it was like, oh my gosh, like one less thing, like one more streamlined thing from like an operational perspective. And so I just thought that was like, I don't know, the cherry on the top.

That's great for me.

So I'm sure there's a lot of DesignOps folks who also feel that as well. I want to talk about— I think what everybody wants to talk about is generative AI and, you know, what the possibilities are. I think there's this, you know, there's this conversation about generative AI, but there's also this kind of this conversation. There's this fear right now of is generative AI going to take over my job? And I think a lot of designers are feeling that. Design ops people, content people, you name the gamut. I'm curious what your thoughts are on that.

Yeah, this is, this is a topic I'm sure everyone's been asking themselves and asking other people. And we all as leaders get asked a lot. I think there's like two different ways to answer this question. Initially, I was thinking I'll take the logical answer and tell you all the things that, you know, we've seen that generative AI can't do and all the ways in which we need to help guide these technologies and use them. But I actually think it's like, it's more, more interesting to start with kind of the emotional answer, which is just acknowledging that something has like changed in terms of everyone's mental model about how they thought the world works or how they thought the world would work. You know, a lot of people at some point in their life go through what many would consider like a traumatic event where you have to kind of reevaluate everything you've been building your mental model around because something just radically disrupts it and makes you kind of, you know, almost lost and anxious and sometimes even depressed. And I think on some level, the entire world has been going through that for the last couple of years because this technology that we thought was akin to magic is now starting to become a reality. And I think it's natural to, like, you know, in this situation to start to, like, wonder, you know, what are all the bad things that could potentially happen? And as a result, it's really easy to kind of not look at all the amazing things that are going to happen. And I think Rafik's presentation was just a tremendous example of this. I think that there, especially for designers and people who build software, there's, there's this incredible moment where we now have these new input modalities that we've always kind of aspired to be able to incorporate into our software that are finally actually possible. And we're in the very early innings of actually figuring out how software evolves to incorporate these new input modalities. We're basically just testing like one extreme, which is just put everything into a chat interface. But I think at Figma, like, just like all of you, we do not believe a chat interface is the optimal interface for every single product and every single experience.

Yeah.

And I think that there is just so much opportunity to think about how you blend these different input modalities, uh, and these different experiences and redesign software, and also how software can solve problems that it never was able to solve again, and what sorts of new opportunities that's gonna create for every single person in this room and all the businesses were a part of. And so I think that, you know, oftentimes when you go through these events in your life, you start off in a place of anxiety and confusion and sometimes even depression. But great things come out of that because you kind of get this moment where you get to realize that everything is new again and you're kind of like a kid learning all over again. And I think that we're starting to see that happen. And I'm actually incredibly excited about the future that that's going to bring. And we can come back to some of the tactical reasons why I think the technology's not there too.

Yeah.

Yeah.

I'd love to demystify a little bit some of the use cases that you're seeing where you think AI has actually been incredibly complementary.

Yeah, I mean, I think we've all kind of gotten tired of all the cute demos online where AI does something magical, but what you don't see is all the ways in which AI also failed at very similar tasks. And that's not to belittle these new technologies. I personally think they're amazing and create this insane amount of opportunity. But the reality is like a lot of people are trying to pick these extreme examples of AI doing things that are so disruptive to kind of force people to acknowledge them and play to this anxiety that everyone in the world is feeling right now. But the reality is, if you look at a lot of the software and, you know, the demos that have been put out there, they oftentimes have these kind of like traditional startup growth curves where a bunch of people come in, kick the tires, and then not that many people stick around. And so when you think about like the experiences that have actually already proven themselves to be able to kind of last, I think like GitHub Copilot is like a great example of this, right? And I think the reason why it's so powerful is that it's not like trying to promise a way in which you can just totally change the way you think about building software. Instead, it's trying to be there right when you need it to help automate some of the tedious kind of things that you didn't even really care about doing in the first place. But it still does it in a way where it's, it's, it's giving you not just control, but the ability to kind of adapt your mental model and keep up with it and really reason about what it's doing as sort of an assistant rather than a replacement. And I think that's like really, really important because even if ultimately we don't need to like be involved in all the details all the time, I think understanding the details is what gives us the creative energy to really reimagine the system and think about how to bring something to the next level. And so I think regardless of whether or not details can be automated, I think we're all going to still enjoy looking at them and understanding them.

Yeah. So this might be a little bit of a spicy question, but what about like the wide range of digital things that get designed? Like, and what's GenAI's role in that and Why I'm asking this question is I think that AI can be a partner, like not the solution of things, but AI can't understand emotion, right? AI can help with things, but like what is the creative impact of AI? And I think the reason why I'm asking that is I think, you know, coming from the design world, like we're so used to doing like, you know, research interviews and stuff like that. Do you think that stuff's going to go away or do you think AI is going to like help complement research? And help complement the things that we're doing to make the user experience better.

Yeah, I think that like the thing we always have to kind of remember and ground ourselves in is that we're building these technologies to serve people, right? And so even if there is some world in which we have AGI and in theory these machines can make decisions independent of people, you know, and all that kind of dark stuff, I think the reality is the reason we're all excited about these technologies or interested in learning them is not because we want to replace people, it's because we want to use them to better cater to the needs of people. Yeah. And I think like at the end of the day, we are people ourselves too. And like these models, they get trained to essentially compress the world's information and then they get trained again to sort of decide what they do with that information. And we're a part of that. Like we decide what they get trained on. We could decide how to use them. We decide what context to give them. We kind of hold the keys to the kingdom in some sense, because ultimately these models don't understand everything and they don't understand us as people.

Yeah.

And so I think it— I think that there's an opportunity to see these things as more of a paintbrush. I think Rafiq did a great job of defining all of us this morning around this concept.

Yeah.

I mean, I think that's a great, like, very visual example of what this can all become. Yeah.

And all right, so we've— we're talking about a lot of companies putting out AI products. I think there's this kind of race right now. You see AI everywhere you look. Like, literally, like we were just talking in the back, like when you want to like return something online, like there's an AI bot. Like when you want to message your doctor, there's like an AI bot and all of that stuff. And I'm curious, like, do you think that— like, the reason why I'm asking this is, do you think that there's a race to get this done because like investors are wanting to do this, to put these into products for like financial reasons? Or do you think people are trying to get this in because there's a race and they want to compete with everybody else? Or do you think they're putting this in because they think this is a good idea and this is actually going to help people?

Yeah, I think it's a little of everything right now. Right. I think that because there's just so much opportunity to do something that's radically different than what's been done before. Yeah, we're all starting with like the obvious opportunities, like, you know, you're just trying something and, you know, letting the world react to it and seeing what sticks and what doesn't. I think there's also a lot of pressure to show that we're innovating ourselves and that we're keeping up with technology. And so, yeah, you know, in some sense it's not that we're caving to the pressure, but we're actually trying to, you know, capitalize on this moment and keep up and learn at the same pace as everyone else. And so I think, yeah, again, like, I think it just comes back to this idea that like that the reality is the evidence points to we're in the very early innings of this because not all of these experiences are sticking, as we all know. Right. And not all of these experiences are actually better than the ones they're replacing.

Yeah.

I think like everyone in this room has probably used Figma, at least to some extent. And you've probably experienced like the kind of rapid expression of color by dragging a color wheel. I don't think any of us like want to replace a color wheel every day. With a prompt where you have to describe those colors that you're trying to explore. That just doesn't make sense. Like, direct manipulation is here to stay for a lot of different use cases. And I think sometimes when you directly manipulate something in a more physical way, it's also a different way of learning and interacting with it and understanding it. And I think like at the end of the day, we all have to understand the software we use and we design and we build.

Yeah, thank you. Okay, next question for you. What are you excited about? With AI? And what are you excited with what Figma is going to do with AI? And if you and I were to come back and talk next year, what do you think we'd be talking about?

Yeah, um, well, at some point, given the title of this talk, I think we got to bring this back to what I do, which is with coding and engineering. Um, so maybe we'll start there and then I'll talk about some other things too. Um, so there's a lot of like, uh, demos online of like taking an image and converting it to code and it's like, oh my gosh, like this is amazing. But like, we look at that a different way. Internally, Figma files are not images. They're already a platform agnostic representation of code. And there's— you don't need AI to mechanically translate them into code. They can be translated into code. That's not the hard part. You can visually reproduce a design in HTML or, you know, whatever you want to do. The hard part is actually inferring the implicit aspects of the design that we as designers or, you know, people who are exploring ideas don't need to express at that moment in time because we're trying to explore extremes and, you know, build consensus and we're leveraging the context of others and we're going to figure those things out later. And so I think there is this opportunity to leverage these technologies to help expedite that inference. So that we as designers and engineers, rather than engineers having to spend the time to go, you know, make their assumption and implement what is basically a mechanical transformation based on an assumption, the software can automate that process, give the feedback to the designer faster, and let the designer iterate from there without having to waste people's time. And I think there's just a lot of opportunity to help accelerate these inferences and validate them more quickly. The other hard part about converting designs to code is the fact that most engineers, or at least engineering leaders, they don't want more code. Like Figma, I'll be honest with you all, we have enough code. We need less code. We need faster code in certain places. We need code that we can all reason about again and stuff like that. And so we're not that excited about a bunch of AI writing a bunch of new code that a lot of engineers don't understand. We want to reuse the code that we're already building to replace some of the older code we have. We want that code to automatically, you know, take advantage of all the accessibility work we're doing and, you know, all the responsive layouts and things like that. And so I also think there's a place for these new technologies to help us better utilize the code we have. And I think in order to do that, engineers have to build a reason about how it's utilizing it and have some control and influence over what it's utilizing. And so, you know, CodeConnect, I think, Even absent AI is already very useful, but you can see it as a sort of building block in terms of, you know, how can we take it a step further and build composite screens in a way that your development team actually wants and you actually want because you're utilizing the components that you've already tested, you've already made accessible, all these sorts of things. So those are the things I'm excited about on the engineering side.

That's awesome. Well, thank you so much for your time. I'm so excited that we had this conversation. I can't wait to come back next year. I have this conversation with you again and see what's happening.

Yeah. Thank you, Meredith. And everyone, I don't know if you've noticed, but Meredith has been everywhere at ZoomFIG. Like, I have no idea how she's maintaining all this context. Let's give her a big round of applause. Thank you. Thank you.

I said thank you. I just said it.

All right.

I think we're done.

All right. Thanks, everybody.

Thanks, guys.

Yeah.