Figma CEO on Design, Product, Engineering: Blurring the Lines in the AI Era

AI just kind of changed the way that we need to think about the surfaces that you're on. Your surfaces are no longer just the owned experience you have when you go to whatever your company name is dot com. It's everywhere people can interact with the product. It's something that I just also really enjoy around AI in particular is kind of testing models and pushing them to their limits. We see improvements not only when we're able to orchestrate around the models better and do smarter things and build more tools around them, but also when the models improve, we improve. We're sort of in this like MS-DOS era of AI where the prompt is everything and the model is some kind of spaceship and your prompt is maybe your compass.

It's finally happening. Dylan, hey, thank you so much for having me and this is an amazing group. I got to meet some folks on my way in and just wow, what an incredible community you've created.

I mean from the start of Figma, it's not like we started and had the Figma community obviously, rather it was the case that there was a design community out there and I was so eager to break into it myself and for Figma. And I was very lucky. Designers and Silicon Valley culture in general is all about paying it forward and folks are so receptive to someone that sends them a cold email and they will actually reply. I try my best to do the same. Not able to always do it, but I do try very hard. And yeah, I benefited so much from not just the ability to meet my heroes in the design community, but then also the chance to really learn from them. I mean, designers give great feedback, especially the really good designers that I was idolizing and reaching out to like a fanboy. And they would sit down with me and just tear my product apart and tell me all the things I got to fix. And early on that was such a blessing, such a gift, and it still is today. So please send your feedback. Email me, over social media, whatever you want to do. We look at it all. But over time, yeah, we were able to also build this Figma community too, both in the product where you can upload resources that can then be remixed and used in other places as long as you attribute, and also in physical space we hold a conference, Config, we're doing Figma community events. We actually just had one last week in the city. And I think there's something magical about bringing folks together in real life.

Yeah, I'm sure I could be better about this. I'm not your resource for time management. And if any of you have good tips, I'm down to hear them later. But I think it's just so necessary. When I can, I try to do like Tuesdays and Thursdays as dedicated product days. Sometimes I spill over to other days to make sure that we're going through everything. Sometimes we're on cycles where it's more ad hoc when there's a lot of travel and whatnot, but the ability to just spend time with a team, dig deep into things, and then oftentimes it's nights and weekends that I get to actually go play and build and try things out. But if you don't actually use and try to use it not just in the way that's natural for you, but use it in the way that other people are using it, and you have to understand that, then you don't actually build the intuition you need to give the right feedback to the team. And I think it's something that I also really enjoy around AI in particular is testing models and pushing them to their limits. So one thing that I do a lot is just when a new model drops, I really like jailbreaking them and getting them to say weird things. I don't know why, but it's just fun for me. And then I send feedback to the labs and they're all like, 'Oh, thank you. That's strange.' I don't know, maybe I'm not like other people. But I learn a lot from that. You know, by getting models into these different headspaces and different places, there are lessons to learn that you can then apply to prompting for professional work, too. And the more you can understand how to structure a prompt correctly or get a model to follow a process with reasoning and thinking, I think that's very powerful.

And as we develop Figma Make, which was referenced earlier, and I saw some hands go up so excited that folks in the audience are using that, and for those that are not raising their hands yet, we'll convert you. But yeah, that's been something that's been so fun to build. And we see improvements not only when we're able to orchestrate around the models better and do smarter things and build more tools around them, but also when the models improve, we improve. That's going to be true across the business, but for Make in particular it's just stark sometimes how much model improvement can just make that product better. And also the prompting strategies that we can help users create or understand or lean into, those are opportunities as well. And we ultimately want to get to a world where Figma Make and Figma Design are two sides of the same coin. Right now you can do things that are more granular edits in Figma Make. You can go point to edit. You can copy from Figma Make into Figma and paste in Figma Design, make edits there, refine, and then bring it right back into Figma Make as part of the prompt. That's really powerful, but that direct connection is the ultimate goal. And there's all sorts of technical specifics about why it takes time to build that out, it's not trivial, but people just want it to work perfectly and that's what they deserve. So we're thinking a lot about how to marry these worlds together of code and design in the right ways where you can get the right benefits in the right places. That's been a really fun journey to be on too, and so much that I get to learn from our team as well about prompting and just AI usage.

I was just in an off-site for design managers like literally an hour and a half ago and was so impressed with some of the work they're doing just to make themselves more efficient. For example, one of our design managers was showing me how he has a workflow where he can spin up a remote agent on a server and from that be able to alert him when tasks are running long or there's some issue in the dependency graph of what we're doing. All these different things that otherwise he has to really crawl through these systems, he can just automate. And so he's got agents that alert him all the time. And yeah, I think there's just so much possibility right now, but it's not evenly distributed yet. We're all learning these workflows.

And even amongst design and PM, I'm really excited for how we can get product managers to be able to share more ideas as part of the design process. Because, you know, as a product manager, you can go and ask your designer, 'Hey, can you make this mockup flow?' and you'll get some version of what you have in your head. Or you yourself can spend 20 minutes doing four explorations in Figma Make and with that explore four different ideas you have, bring them all to the table, and then work with your designer to figure out which one should we refine and how far can we push it, how much can we make this awesome. Because now we're in this world of software where good enough is no longer enough. Good enough is mediocre.

Absolutely. I think that it's been our thesis from the start of Figma, almost the start certainly for the past decade, that design is at the top of the value stack for software. And as you think about that value stack and how it's changing, how the software landscape is changing, there's just an exponential rise of software in the world that we've been on for decades now. I mean, Marc Andreessen wrote an essay I think in 2011 about how software is eating the world. And it's only compounded that exponential curve since then, which is a really wild thing to say. And now we're in this world with AI where software is easier to create, so with that curve going vertical instead of just exponential, what does that mean? How do you win? How do you differentiate? How do you actually put out a product that's going to be the best it possibly can be? And I think it's design, it's craft, it's point of view. It's also your brand, your marketing. And it all has to connect and work together. You need to understand as a product manager or designer, the constraints you're living under, what are the business constraints, what are the business objectives, what is the way the brand needs to operate, what's our point of view on our space, our world, and all of that, plus the culture of the moment plus the specific flows you're working through. It all connects together. And if you're not able to internalize that all and think through it as a system and then articulate in a way that's joyful and wonderful for the user, I don't think you're going to win. And because of that, because design is how you win or lose, we need everyone in that product development process to care about design. And I also believe that with tools like Figma Make and the ability to use a design system you've built internally in Figma Make, you're able to make it so that what you create, the idea or four ideas that you put out there to explore that option space, is more consistent with the way that things actually look and work. It's not going to be the highest craft thing that your designer will work on and create with you, but it will be something that passes that's credible. And if you can get to credibility, then someone will actually look at that idea, not just dismiss it because it doesn't look right, but evaluate it on its own terms. That's a good starting point.

Correct.

Well, I think it's interesting the word subjective even, because I think that's the beauty of design, that it's non-deterministic. It's not obvious, and that's the art and science of it. It's the opportunity. If it was all just objective and deterministic and clear, we would all be making the same things all the time. It'd be pretty boring and you'd have no way to differentiate yourself. The possibility to go push the boundaries with design and actually lean into the craft, wherever you're coming from in the org, people should be thinking about this and figuring out what's the strategy to do this. That's immense and it can inflect your company literally. The distribution of what we're seeing in the market right now is all over the place. You've got folks who are just trying to figure out how to build their design team, what's a design system, best practices, all the way to literally people that are mirroring, executives and CEOs that are basically going into talent wars for design just like we're seeing with AI research.

And I think that end of the distribution where people are valuing design so much and making the designers they bring in key leaders alongside product, really making sure the folks that are advocating for great craft are the ones that are able to shape the roadmap, shape the future of the product. Those are the founders and CEOs that I am very bullish on. I know who they are and they're the folks that I legitimately think will win. I think their competition's behind. So it's really important as a product leader, as a CEO, whoever you are, to be advocating for the voice of design regardless of where that voice is coming from. Otherwise, I think you're in a rough shape. And if that's not the org you're in right now, if there are constraints working against you, I think you got to go fix it or maybe find another org.

Yeah, we started with FigJam as an app. So you can now invoke the Figma app through ChatGPT. And I think this is pretty exciting because first of all, ecosystem. For some folks it's not as intuitive yet. We're all kind of learning what this new ecosystem is going to look like. For me, my intuition is that AI just kind of changed the way that we need to think about the surfaces that you're on as a software business or company. And your surfaces are no longer just the owned experience you have when you go to whatever your company name is dot com. It's everywhere people can interact with the product, and you should be broadening the scope of where people can interact as much as you can. So what does that mean for us? First of all, we launched our MCP. The initial use cases are obvious. You're in Figma Design, you want to go build the thing. We've had Dev Mode for a while. Now we've got Dev Mode MCP. If your file is well structured, it's got auto layout, it's got variables, you click on something and pull it in from the MCP and you can build it much faster on the front end. You still need to work on the craft and make sure the interactions are right, there are a lot of details to work through, but the speed at which you're able to move is fundamentally much faster. It's been amazing to see the reactions to that. Then it's like, okay, let's go further. So in Make, yesterday we launched connectors to make it so you can pull in your Notion data, your Atlassian and Jira data, Linear, more. It's technically launching in a few weeks, we announced yesterday, so coming soon. But I think that's just the start.

As you're able to pull in more of this data from all these different places, it's objectively just helpful when you're in the creation process. With ChatGPT, now we're in another surface invoking. And for me, this is just so obvious that of course you want to make it so that wherever you are, in whatever context you've got built up, you're able to then use that context to go and convert it into whatever artifact you're trying to create. And this creates a very happy path for the user. Then you can say, 'Okay, make this a diagram.' And it pulls up Figma, FigJam in particular, and it can use that context to go create a diagram. You can then go back and forth. You can prompt FigJam. You can iterate on things. If you want to go further, you can open it in Figma on our platform, save it to your account, share it out, or visually tweak it with all the tools we provide. Depending on your use case and who you are and what you're trying to accomplish, you might be doing any of those things. Maybe you never see the data, maybe you go and do extended editing. Maybe you're a Figma user, maybe you're not. It doesn't matter. We're trying to make our tools more broadly available and pull more people into the broader ecosystem, whether it's on figma.com or elsewhere. And I just think it's super key that people change their thinking around this and adjust to a world where the surfaces you appear on are maybe everywhere. So we're very open right now on the partnership side, trying to do as many partnerships and integrations as we can as long as it's the right setup and the right user experience, because we think this is the future.