Figma launches Design Agent and Dylan Field explains why designers shouldn't outsource their thinking to AI

Speaker 2: Let me tell you about CrowdStrike. Your business is AI, their business is securing it. CrowdStrike secures AI and stops breaches. We got Dylan Field, the CEO of a public company coming on.

Speaker 1: We've been keeping him waiting.

Speaker 2: We've keeping him waiting. We're very excited to have Dylan Field from Prisma, of course. Welcome back to the show.

Speaker 1: Here we go. Hello.

Speaker 2: We were having a lot of fun with the head of the National Christmas Tree Association.

Speaker 1: I heard part of it.

Speaker 7: And it was funny because I was just thinking that Christmas came early. We launched Figma Design Agent today. So...

Speaker 2: Oh. There you go. That was fantastic.

Speaker 1: On that, I think that's appropriate Okay. To to

Speaker 2: Yes. I gonna... Presents are under the tree. Okay. I think we have enough we have enough Christmas nonsense. Tell us more about the product launch. Take us a click deeper, please.

Speaker 7: I mean, it's been awesome. We've been working on this for a while. Mhmm. We've had the beta going and just been hill climbing. The team has done an incredible job

Speaker 2: Yeah.

Speaker 7: Of not just working with our own first party models

Speaker 2: Yeah.

Speaker 7: Frontier models, but also really thinking about how do we make this great for our users.

Speaker 2: Yeah.

Speaker 7: And so much more to do. But at today's launch, you'll see that the Figma agent's faster, higher quality. It has better ability to deal with context across your files Yep. And more overall agentic capability. Yeah. I would say it's improved quite a bit on design systems. We still have more room to go on all the dimensions, definitely design systems too. Yeah. But really proud of the launch and find myself really enjoying using the Figma agent on the Canvas. And I think as we think about just kinda where it all goes, there's so much that you can do in the Canvas. It's not the case that, you know, the nondevelopers, the designers, the PMs feel at home in the terminal and the IDE. I don't think text is the final representation of everything. I love text. I don't think that that's it. And I believe that we're going to find more ways to go build software and use these amazing reasoning models to be part of that process, not the entire thing.

Speaker 2: Are you

Speaker 7: It's humans agents together. You know?

Speaker 2: Are are are you generally through this process, yeah, I mean you're building an agent that is going to excel at one discipline and not be a AGI model, right? And so I'm wondering if you have a broader philosophy around ensembles of models. I mean we've seen the decisions API, JEV. There's like smaller models for different purposes. And you can imagine this future where you're actually using more models, more models. I've been for a long time I was like, actually let's get rid of the model picker. Just give me one thing. But and maybe that happens and there's an abstraction layer on top of some of these tools. But it does feel like there's a ton of value in having a team, research scientists, AI researchers focused on a particular discipline, shine the eye of a sore on on something and go and deliver a great product that then can be the right tool for the job. But are you in a world where you're applying your lessons here to like the shape of AI as a tool broadly, or is this a unique design thing?

Speaker 7: Superintelligence. It's, you know, it's the goal. No. I'm joking. No. And we're we're trying to make, like, the best design agent Yeah. To power designers. We are not trying to replace designers. We're trying to make it so they are maximally empowered

Speaker 2: Yeah.

Speaker 7: In their work, which does mean

Speaker 6: Yeah.

Speaker 7: Helping them make PRs.

Speaker 2: Yeah.

Speaker 7: Helping them go even further with their work, helping them get all the context they need to go make decisions. Yeah. And it does so happen that, like, after all this time building Figma, the team really does understand the needs of designers

Speaker 2: Yeah.

Speaker 7: And what we need to do to go achieve that. And I think that is just as much a part of the... A special sauce over time. Yeah. And so, yeah, I'm very, very excited about what we can do here and also how the the loops that we have in the product and the data available Yeah. To be able to work with in addition to just kind of, like, this very, you know, singular focus of how to make the best thing for design. Yeah. And the best thing for designers to use every day and make them more efficient.

Speaker 2: Yeah.

Speaker 7: Like, I think that is really nice to have that focusing power. And also, you know, you don't wanna be in this mindset of you're fighting against bitter lesson. Yep. Want to get the advantages of the general models, the ways they've been trained

Speaker 2: Sure.

Speaker 7: On all sorts of things, associations they've made, and you want to make it so that you're really good at design.

Speaker 2: Yeah. No.

Speaker 6: It makes a lot of sense.

Speaker 1: Yeah. I mean, it feels like part of the challenge as a design company is is you don't have the luxury of being able to produce slop even at at anywhere this... At the same rate as some of these other companies where, like, they might be offering a bunch of things and they're not like... Their whole brand is not built around like design quality and and taste and and, you know, precision and so they can just make something that like only 5% of the time it's actually like decent and and that's enough. Whereas I think for... So I guess like, do you feel like do you feel like you have to hold the team to a higher bar on... In terms of what products are shippable? Because like a new y c company can be like, hey, we made a design agent. Like, you should try it. And if people even have like an okay time with it, they'll they'll be, like, more excited about it. Whereas your community, again, has, like, much higher than average taste level, capabilities, all these different things.

Speaker 7: I think there's a bunch of points in there, and it's like, oh, they're all interesting. The... Maybe we'll start with just the overall setup that we have and what we're really caring about for this product, which is we don't... Again, we're not trying to replace designers. We're trying to empower them. And as we think about that, what does that mean? Well, it means that as you're trying to get something in your head onto the canvas, we've talked this before. Mhmm. You might be aided by prompting. And, also, it might be the case that that's just the starting point. You might have to do more work, and that requires more integration with a global system of Figma. And so you kinda have to teach the models Figma first and make it so that they can understand how to go and work with Figma as well as do good design work. But I think that right now, you'll probably see, you know, just as much on the slop side across all products in, you know, in AI. I hope you'll see less in Figma. Mhmm. We are working to make it so that there's less and less and less and and better and higher quality results. But, also, it's it's... Taste is in a a wild thing in that different people in in different domains have different definitions. And so the ultimate is how do we make sure that we can understand what you're going for and give it to you.

Speaker 2: Mhmm. How are you thinking about interfacing with AI as a top of funnel? I think everyone saw the blender moment when everyone was modeling everything in their world in blender and the you know the AI community and and and AI enthusiast community was downloading and installing blender and it and it feels like there's there's immense value if you can be, you know, a button there or even just the the the top of funnel agent says, hey, you you've been working on this business idea. I did a first draft of a logo but you should take this much further. Why don't I hand off to Figma, get you set up there, even pass along credit card details? Like there could be a really deep integration but then there's always difficulties on where the value accrues and whatnot. How have you been how you've been dealing with those potential trade offs in, like, an increasingly AI agentic top of funnel world?

Speaker 7: Well, first and foremost, we wanna always be learning from the ecosystem.

Speaker 6: Mhmm.

Speaker 1: And I

Speaker 7: think that's essential. The ecosystem sometimes in these moments evolves so quickly

Speaker 2: Yeah.

Speaker 7: That when you start to see everything come in, it's like, wow. There are so many ways that we can work with other companies or other models or different setups here. You know, sometimes it's the case where you're like, wow. Yeah. I predicted that, like, five years ago, and other times, it's more of, I did not predict this at all, and this is a second or third order effect. Woah. Yeah. And I think that if you don't understand what people are trying to do, you can't build for them. Mhmm. And, also, you need to work well with the ecosystem in order to make all these new routes possible. And there's complex questions that, like you mentioned, around where does value live, where does it accrue. I think that some of those, there's a category of questions where it's... There there are not clear answers, and it's not clear there will be clear answers. There's other categories where I think that actually the way the space evolve... Will evolve is more clear. And so you also have to have, I think, some macro bets about, like, the way things will play out over time, but not get stuck in your head.

Speaker 2: How are you thinking how are you thinking about motion graphics? I feel like we've seen some cool, like, one shotted demos. I've done enough motion graphics in After Effects to know that, like, does need to be pixel perfect and sometimes you need to, like, make small changes and doing that via text would be extremely cumbersome. And so it feels like there's an increasing opportunity there for web based, node based AI enabled, MCP enabled tooling. What's the current state of the strategy there? How do you see the motion graphics discipline evolving?

Speaker 7: I I think that the way that we'll see motion graphics evolving is parallel to all these other areas too, which is when you're facing a blank canvas or a blank page, it is hard to get started. Yeah. The sooner that you can kind of build to think Yeah. Get something out onto the canvas, whether that's through design and... Or it's even in a sketchbook or you're doing something with a coding model. Yeah. That is good. It helps you get over that initial hurdle, and then from there, you can react to it and go, okay. What do I wanna tweak? What do I wanna change? How do I make this mine, and how do I make this my vision? And software ultimately is just amazing creative medium, and we don't want the center of the distribution for everything. We don't want everything to converge to the mean. Rio, who was previously a cursor and recently left after they were bought by SpaceX.

Speaker 3: Yeah.

Speaker 7: He had this amazing post recently where he talked about some of these ideas. And I think that Weave is another area for Figma where we're really trying to figure out how to help people do this, both use models, get all the power from them, but still apply their own judgment, creativity, taste, and really shape it, the model outputs like clay rather than just trying to go for the one shot. The one shot's rarely where you wanna land and end up.

Speaker 2: Give us your design review of the new American passport. Do you like the previous version or the current version more? Which which passport speaks to you?

Speaker 7: The new version, I really like. I think that, you know, beautiful cover and then I love the way that as you flip through it, the eagle animates. We can get

Speaker 2: to that. Yeah. There's a flipbook involved which is...

Speaker 7: I feel someone's blowing something.

Speaker 2: No. No. No. I did I did see that but we didn't... Scoob. Were just looking at the front and the front seemed like a very a very moderate iteration. There was some debate on our team about the kerning on the word passport. Tyler's not a fan of it.

Speaker 7: It is too controversial for me. Let's go back

Speaker 2: to AI. Yep.

Speaker 1: Where do you feels like people on the timeline have been doing Figma's marketing for you in some ways where every other day there's somebody like, if you send me an AI generated slide deck, I won't even take the meeting.

Speaker 2: Oh, yeah. I'm... Said that?

Speaker 1: No, I've been saying that for a time. I've been saying that for a long time. Just because I think you should be thinking about, you know, if there's a sea of just sameness Yeah. Being able to prove to people that you can think differently, stand out, have a unique point of view, opinion, all these different things, you know, I would... In a sea of AI generated decks, would almost rather see like a Google's clear obvious Google Slides template because that is signaling, hey, I I I don't I don't actually care about design at this very moment, but I care about showing that I communicated, you know Words. Very clearly what I'm trying to do. And and the problem with AI generated decks is that every single slide, I'm second guessing how deeply did you think about what you're communicating.

Speaker 6: Mhmm.

Speaker 1: Because I don't wanna I don't wanna I don't wanna spend time trying to process and evaluate somebody in their thinking if it's possible they weren't thinking at all.

Speaker 7: I I think that's a great way to put it, and have really been kinda on this thread myself with, our team. You know, I'm... At some point recently, I kinda gave some guidance to the team around AI writing, for example. And I had thought about it a lot. It was a post that I kinda had queued for, I don't know, three weeks or something, I kept rewriting it, ironically. Because I thought this is really important for me to... Like, we can't just ban AI writing in the company. Then we're not empathizing with, like, everyone else that's in tech Sure. And everyone else that's, like, using our software. At the same time, I think that there's degradation of trust that can happen as you have all these folks interacting across the company and people are second guessing, did you even read what you're asking me

Speaker 2: to read?

Speaker 7: Mhmm. Did you think through this? And because of that, I I came up with this old sort of rule for Figma, which is... Let's just have it be that you put written with AI at the top of messages Oh, yeah. Where there's AI writing of any kind. And, you know, if you're using AI to, like, spell check, essentially, proofread, and you change a word, I don't care. If you write the whole thing with AI or there's, like, good paragraph in there, yeah, just put it right with AI. It's easy. Yeah. No big deal. And that way, we can trust that if you're sending something, you care about it, and you've thought through it.

Speaker 2: Yeah.

Speaker 7: And I think that in general, people should not give up on their own thinking. They cannot give that to the model. And I see so many people that are essentially looking for the shortcut, trying to make it so that they can have... Maybe less than Figma, more outside of Figma, but folks are eager for the model to automate for them, for them... The model to think for them. And I think that the more important thing you can do is to actually use the model to learn, to try things you wouldn't have tried before

Speaker 6: Mhmm.

Speaker 7: To be more daring, and to kind of expand possibilities rather than contract them.

Speaker 1: Yeah.

Speaker 7: And I don't think... I think it's good to have default skepticism, not because there's models are all bad or something. That's not what I'm saying. But rather because if you're default skeptical, then you will understand better how models work, what to use them for, and what not to use them for, and you'll get better results. So, yeah, I'm all about thinking.

Speaker 1: Mhmm. Yeah. The... I've noticed in in the process of making decks, like my best decks always, the outline took like twenty minutes max, where I just like actually had clarified. Like you wanna... Mhmm. There's there's increased clarity that comes from trying to communicate an idea or a company or a product or a launch or a campaign. Like, you're gonna naturally get more clarity on the topic as you explain it in a way that is sort of logical and and has a consistent narrative. And then there's times that I've opened this sort of my, like, deck outline in Figma that I've shared at this point with, like, over probably a 100 founders. Right? Because I spent so much time on this. I have example slides for each slide and, you know, structuring and all this stuff. And there's been times I've landed in that and I've just... I'm just wrestling with it for hours and hours and hours, not making progress. And and two things are usually happening. One, it's just a bad idea. And... Or or two, I don't understand it. I just don't understand it well enough, or there's just like something fundamentally broken. And so when people go to AI first for something as important as like communicating your company, like, you're you're robbing yourself of like one that that sort of check and balance of do I actually understand this deeply? Does the thing that I'm presenting make sense? And you're sort of accelerating to the next step, which is pitching the thing. And the the last thing you wanna go do is pitch an idea to a bunch of smart people that is not well thought through, that doesn't make sense. And and so you're, like, saving that time of, like, wrestling with the idea, but then wasting all this... The the... Wasting even more time and wasting other people's time trying to communicate something that doesn't make sense. And so...

Speaker 7: I mean, if you can't explain something, don't know it yet. Yeah. I think it's the basic rule I always have gone by, and that's pre AI too.

Speaker 2: Well, I don't know it, but my assistant knows it. And so you can just talk to the model now. You don't even need to talk to me.

Speaker 7: There's also the flip side of, like, there's also the flip side of of not just the do you understand this, can you explain it, but also, like, what are the things that no one knows yet that we all need to explore and, you know, that can't be fully understood. And I think that mystery in the world and sort of, like, the the ways that you feel just when you encounter something that's that's really novel, new, or or as a angle on something that exists but just is... It's fresh. Those are magic moments, and and I think that, in general, there's an opportunity right now to lean more into mystery in a time when everything's becoming a bit less mysterious.

Speaker 2: Yeah. That's interesting.

Speaker 1: Mic drop.

Speaker 2: I love it. Thank you

Speaker 1: so That's much a good place calling to end because I have follow-up questions to that encouragement of mystery, But you can just leave it in the sort of dramatic pause. Yeah, you'll

Speaker 7: leave it mysterious.

Speaker 1: Leave it. That is mysterious. Thank you so much. Great to see you,