Tech Disruptors

Kris Rasmussen on Figma's AI Strategy: Design Context as a Moat and Hybrid Pricing Models

42 min Listen to the original
Artificial IntelligenceSaaS PricingEnterprise SoftwareDesign SystemsAI AgentsModel Fine-tuningHuman-AI Collaboration

Summary

Kris Rasmussen, Chief Architect at Figma, discusses how AI agents serve as interface glue rather than replacements for traditional UIs, emphasizing that design systems provide critical context for better AI outputs. He argues that as code becomes commoditized, value shifts to design, with Figma leveraging proprietary user data to fine-tune open-weight models for specific workflows. Rasmussen highlights bottlenecks in AI adoption due to security and backend expertise gaps, noting that human curation remains essential. He also outlines Figma's infrastructure strategy, which avoids building foundation models from scratch, and predicts a shift toward hybrid seat-based and consumption-based pricing models.

Key takeaways

  1. 1
    Figma’s user base is split equally among designers, developers, and other collaborators, indicating the platform has transcended its original design-only identity to become a collaborative infrastructure tool for entire product teams.

    Kris Rasmussen · 2:07

  2. 2
    The speaker frames AI agents not as replacements for traditional software interfaces, but as the 'glue' that connects disparate tools, suggesting that software companies that maintain strong, specialized UIs will retain relevance by integrating with agentic layers rather than being bypassed.

    Kris Rasmussen · 10:17

Episode sections

  1. 0:19 – 1:47

    Podcast Intro and Guest Background

    Anurag Rana introduces the Tech Disruptors Podcast and guest Kris Rasmussen, Chief Architect at Figma. Rasmussen outlines his 10-year tenure at Figma, previously as CTO, and prior roles at Microsoft, Asana, and Aptana, noting his recent transition to focus on AI and agent workflows.

  2. 1:47 – 10:17

    Figma Product Evolution and User Base

    Rasmussen describes Figma as a full-stack creative canvas founded in 2012, serving an equal third of designers, developers, and collaborators. He explains how the platform evolved from a design tool to include FigJam and Slides to support broader team alignment and brainstorming.

  3. 10:17 – 13:32

    AI Agents as Interface Glue

    Rasmussen argues that AI agents act as the 'glue' connecting various software interfaces rather than replacing them entirely. He contends that traditional UIs remain relevant for specific tasks, and the best software combines agentic interfaces with traditional input modalities to leverage user muscle memory.

  4. 13:32 – 18:52

    Workflow: From Prompt to Production

    Rasmussen details a workflow where users can start with a design system or external code, use Figma's agent to generate prototypes, and then curate the output. He emphasizes that agents often produce 'accidental' results, requiring human judgment to refine designs before handing them off to engineering or coding agents.

  5. 18:52 – 20:20

    Context and Design Systems as Value Drivers

    Rasmussen states that as code becomes commoditized, value shifts to design. He highlights Figma's advantage in providing rich context through existing design systems, which helps agents produce better, more aligned results compared to linear coding agents that lack this collaborative, visual context.

  6. 20:20 – 22:55

    Enterprise Adoption and Mind Share

    Addressing concerns about mind share shifting to front-end agents, Rasmussen asserts Figma retains mind share by earning the right to build superior workflows. He notes that while integrations with Slack and Teams are important, Figma remains the central place for alignment and context creation.

  7. 22:55 – 29:37

    Bottlenecks in AI Adoption

    Rasmussen identifies a key bottleneck: the assumption that AI prototypes can ship directly to production. He explains that designers often lack the domain expertise for security and backend systems, leading to 'accidental' code. Consequently, teams still use Figma to curate context and align teams before implementation.

  8. 29:37 – 30:52

    Infrastructure and First-Party Models

    Rasmussen reveals Figma directs more internal traffic to its first-party models for specific subtasks within the Figma Agent product, which recently reached general availability. He notes the infrastructure has evolved to support training loops for these models to deliver results faster and cheaper.

  9. 30:52 – 31:56

    Figma's Model Strategy and Infrastructure

    Kris Rasmussen explains that Figma does not build foundation models from scratch but leverages open-weight models, optimizing them for specific design workflows rather than general-purpose use. This approach allows for more efficient training with less capital compared to major labs.

  10. 31:56 – 33:19

    Model Agnosticism and Cost/Quality Trade-offs

    Figma employs a model-agnostic infrastructure that selects the best model for specific tasks, such as using frontier models for interactive prototypes and different image models for raster imagery. The strategy prioritizes quality first, then optimizes for speed and cost, while also fine-tuning open-weight models using proprietary data.

  11. 33:19 – 34:42

    Competitive Advantage via Proprietary Data

    Rasmussen argues that Figma's advantage lies in its vast quantities of design data and user trajectories, which allow for fine-tuning models to be better at design and specific to Figma's ecosystem. This data access is something external AI labs do not have, enabling Figma to improve efficiency and reduce errors.

  12. 34:42 – 36:01

    Human Alignment and Design as a Moat

    The speaker identifies aligning AI outputs with human taste and judgment as the hardest problem in AI, positioning Figma's decade of experience in curation and context as a key differentiator. He asserts that while models are becoming commoditized, the tooling and workflow for humans remain a complex domain where Figma excels.

  13. 36:01 – 38:50

    Token Management and Internal Culture

    Figma has not deployed hard caps on token usage internally to foster experimentation, though they are exploring soft caps. They manage customer token spend by providing controls, fixed credits per seat, and highlighting best use cases in internal meetings to educate users on efficient model usage.

  14. 38:50 – 42:13

    Pricing Evolution and Customer Willingness to Pay

    Rasmussen predicts a shift from pure seat-based SaaS pricing to hybrid models combining seat-based and consumption-based components. He notes that customers have shown an eagerness to pay for outcomes and better results, provided the pricing remains competitive.

Claims worth verifying

Figma was founded in 2012.

Unidentified speaker (UNKNOWN) ·1:47

How to verify: Figma corporate history page or SEC filings if public; company press releases. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Primary source

Approximately one-third of Figma's active user base are designers, one-third are developers, and one-third are other collaborators.

Kris Rasmussen ·2:07

How to verify: Figma company blog posts, earnings calls (if public), or official press releases detailing user demographics. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Kris Rasmussen served as CTO at Figma before becoming Chief Architect.

Unidentified speaker (UNKNOWN) ·1:47

How to verify: LinkedIn profile of Kris Rasmussen or Figma leadership announcements. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Primary source

FigJam was introduced during the pandemic.

Kris Rasmussen ·7:15

How to verify: Figma product launch announcements or blog posts from 2026-2026. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Figma recently launched its Agent product to General Availability (GA).

Kris Rasmussen ·30:12

How to verify: Figma product update announcements or release notes for the Figma Agent. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Figma directs more internal traffic to its first-party models for specific subtasks.

Kris Rasmussen ·30:12

How to verify: Figma engineering blog posts or technical documentation regarding their AI infrastructure. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Kris Rasmussen was an engineering lead at Asana.

Kris Rasmussen ·2:26

How to verify: LinkedIn profile of Kris Rasmussen or Asana press releases. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Kris Rasmussen was chief architect at Aptana.

Kris Rasmussen ·2:26

How to verify: LinkedIn profile of Kris Rasmussen or Aptana historical records. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Kris Rasmussen co-founded RivalSoft.

Anurag Rana ·1:18

How to verify: LinkedIn profile of Kris Rasmussen or business registration records for RivalSoft. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Figma has integrations with Slack and Teams.

Kris Rasmussen ·21:59

How to verify: Figma integration marketplace or documentation. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Figma is not building its own foundation models from scratch but is benefiting from open-weight models that have gained steam recently.

Kris Rasmussen ·31:15

How to verify: Check Figma's official engineering blog or technical documentation regarding their AI model architecture and partnerships with open-weight model providers. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Figma optimizes its models for specific user workflows and trajectories rather than building general-purpose models.

Kris Rasmussen ·31:35

How to verify: Review Figma's product documentation for 'Figma Make' or 'Figma AI' to see if features are tailored to design-specific tasks versus general AI capabilities. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Figma uses a model-agnostic infrastructure, utilizing frontier models for interactive prototypes and different image models for raster imagery.

Kris Rasmussen ·32:16

How to verify: Examine Figma's technical documentation or API references for 'Figma Make' to identify the specific models or providers used for prototype generation versus image generation. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Figma possesses vast quantities of design data and user trajectories that external AI labs do not have access to.

Kris Rasmussen ·33:30

How to verify: This is a proprietary claim; verify by checking Figma's privacy policy and data usage terms to see if user design data is used for model training, and compare with public statements from major AI labs about their training data sources. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Figma has not deployed hard caps on token usage internally to foster experimentation.

Kris Rasmussen ·36:12

How to verify: This is an internal operational claim; verify by checking Figma's internal employee communications, if available, or industry reports on AI token consumption policies at design software companies. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Figma provides customers with controls over spend, including fixed credits per seat and the ability to purchase additional add-on credits.

Kris Rasmussen ·36:47

How to verify: Check Figma's current pricing page or enterprise billing documentation to confirm the availability of token/credit-based billing options and seat-based credits. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Figma offers an MCP server to allow customers to use third-party agents with Figma's software.

Kris Rasmussen ·38:05

How to verify: Search for Figma's official documentation on 'Model Context Protocol' (MCP) integration or developer resources to confirm the existence and functionality of their MCP server. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

A sizable percentage of Figma's user base uses both internal Figma agents and external third-party coding agents.

Kris Rasmussen ·39:49

How to verify: This is a specific usage statistic; verify by checking Figma's public user engagement reports, blog posts, or case studies that quantify AI feature adoption rates. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Customers have shown an eagerness to pay for outcomes, leading to a hybrid pricing model of seat-based and consumption-based components.

Kris Rasmussen ·39:26

How to verify: Review Figma's recent pricing changes or earnings call transcripts (if public) or industry SaaS pricing trends reports to confirm the shift towards consumption-based pricing in the design software sector. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Kris Rasmussen is the chief architect at Figma.

Unidentified speaker (speaker_2) ·30:52

How to verify: Check Figma's official leadership page or LinkedIn profile for Kris Rasmussen to confirm his title and role at the company. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Do your own research

  • Verify Figma's current pricing structure and any announced shifts to consumption-based models via their official pricing page or recent earnings calls.
  • Review Figma's public documentation on AI agent capabilities and limitations to assess claims about 'accidental' outputs and security risks.
  • Analyze Figma's competitive position against other design tools (e.g., Sketch, Adobe XD) and emerging AI-native design platforms.
  • Check for any disclosed conflicts of interest or partnerships between Figma and major AI model providers mentioned in the transcript.
  • Research the technical feasibility of 'round-tripping' between Figma and third-party coding agents via MCP servers as described.
  • Evaluate the claim that design data is a unique moat by comparing Figma's data access to that of general-purpose AI labs.
  • Look for industry reports on AI adoption bottlenecks in enterprise design workflows to validate the speaker's assertions about security and backend expertise gaps.
Full transcriptexpand
0:19
Anurag Rana: Hi everyone, welcome to another episode of the Tech Disruptors Podcast, brought to you by Bloomberg Intelligence. Bloomberg Intelligence is the investment research platform on the Bloomberg Terminal. We have 500 analysts and strategists working across the globe, and we are focused on all major world markets. Our coverage includes over two thousand
0:38
Anurag Rana: equities and credits, and we have outlooks on more than ninety industries and hundred market indices, currencies, and commodities. I'm Anura Anuragrana, Grana, a technology analyst at Bloomberg Intelligence, and I'll be your host. for today's episode. We're delighted to have Chris Rasmussen as our guest today. Chris has been with Figma for
0:58
Anurag Rana: almost 10 years, where he's now the chief architect driving the company's technology vision, helping solve engineering challenges. For most of his career at Figma, he was the chief technology officer building out the tech organization. Prior to that, Chris served as an engineering lead at Asana, as a co-founder at
1:18
Anurag Rana: RivalSoft, and as chief architect at Aptana. Chris, welcome to the podcast.
1:23
Kris Rasmussen: Yeah, thank you for having me. I'm excited to be here.
1:25
Anurag Rana: Our podcast today will focus on how AI is changing, how users design and create new digital products.
1:32
Unidentified speaker (UNKNOWN): So,
1:33
Anurag Rana: Chris, for some of us who may not know Figma a lot, um, please give us a bit of a background on the company, the products, and more importantly, also about your background and how you came about to Figma.
1:47
Unidentified speaker (UNKNOWN): Yeah,
1:47
Kris Rasmussen: I'll start with Figma and then I can go back to my background. But Figma is where teams design and build the world's best digital products and experiences. Founded in 2012, our full stack creative canvas brings together teams, agents, code, and design all in one place so that companies can go from idea to ship product faster. Uh, we began as a design
2:07
Kris Rasmussen: tool, but it's important to understand that approximately one-third of our active user base are designers, one third are developers, and one third are other collaborators that they work alongside. Yeah. So before Figma, as you mentioned, I was uh uh I had I had a number of different roles, leading engineering teams and building uh real-time infrastructure at companies like Microsoft, a
2:26
Kris Rasmussen: Asana, uh, and Aptana. Um, and uh and also Rivalsoft, which was my own company. And then in the later years, I actually took a little bit of a break to to pursue some passion projects. Uh, and I started developing game and physics engines to to build some cool surface simulators, which is something I've always been passionate about on the side. And I was kind of using that as a pause moment to kind of revisit
2:46
Kris Rasmussen: what I wanted to do next in my career, thinking that I'd probably go into VR or AR. Um, but after looking more closely at the space, I started to wonder if it was really the right market to bet my career on. And so I started doing technical advising and consulting on the side to sort of figure out what was next. Uh and during that time, I actually had the opportunity to work with Figma as a contractor. So I came in and for the eight
3:06
Kris Rasmussen: months, I was pretty much just writing code. Um, but uh during that time I realized it was a really special team and a really special place. And I guess I had a good enough impact. impression that they offered me the the role to to lead engineering and join full time, which I did. And so that's how I joined Sigma. And I thought it was like the perfect intersection of my background with graphics that I'd been doing on the side, but also my professional career
3:26
Kris Rasmussen: building real-time collaborative infrastructure.
3:28
Anurag Rana: That's very interesting. I did not know that, Chris. I was doing some background reading on you as well. So I I I Uh obviously I I didn't find that. But tell me a little bit about it then. Uh you know, when you were looking at these game engines. you know, were people using Figma to design those products? Or, you know, who who would be the, you know, who do you who do you think is your, you know, a typical
3:48
Anurag Rana: designer or a builder on Figma right now?
3:51
Kris Rasmussen: Yeah, we're we're used by uh companies of all shapes and sizes in a variety of different infrastructure i i industries, including like the game industry. Um at the intersection with my background was really more around the technology and kind of underpinnings. So with games and graphics, uh, you're really focused on building real time performant infrastructure. Everything needs to operate like
4:11
Kris Rasmussen: lightning fast. So there's never any kind of glitches or blips when you're interacting with the game. And it's the same thing for a design tool. When when designers are working on a real time cooperative canvas, it has this kind of underpinning of multiplayer, just like games. And you need to build these custom rendering and graphics engines to make sure. that even the largest canvases for the largest teams continue to kind of work
4:31
Kris Rasmussen: smoothly without any hiccups. Um it was really fun for me to be able to combine that sort of side passion of of graphics and games with the uh the professional background I had around building these kind of real-time collaborative tools for developers and teams at companies like uh Uptana Abtiana and Asana.
4:47
Anurag Rana: And um, you know, you recently changed roles uh and you're not now the the chief architect at Figma from being the CTO. You know, what what really was the reason for you to shift these roles?
4:58
Unidentified speaker (UNKNOWN): Yeah,
4:58
Kris Rasmussen: I've been in uh I was in the sort of head of engineering role for most of my time at Figma, I guess all of my time at Figma until recently. And um over that time, our product suite and company has expanded pretty dramatically. And it's it's been a really, really exciting change. It's kept my role interesting throughout. Um, but as the team has expanded, we brought on some really incredible leaders underneath me. And
5:18
Kris Rasmussen: at a certain point, I just realized that my domain expertise uh from being at Figma for so long was sort of underutilized relative to the depth I could apply to some of our more most important initiatives. And so I started talking to Dylan about a transition in my role that enabled me to kind of free up some time to go deeper into some of these new spaces that have been emerging around.
5:39
Kris Rasmussen: uh the way that our customers are using Figma alongside agents inside and outside our product in order to um change the way that they design and develop as a team.
5:48
Unidentified speaker (speaker_2): So perhaps we can start off with the Figma, the product itself, go back in the pre AI era.
5:53
Anurag Rana: So if I'm a designer of a new app, or if I have a team, let's say I work in a bank, how would I go and use Figma? in those days? I'm assuming, you know, being a web platform, you had multiple things you can go from one to the other. So so give us a workflow as to what would be my uh
6:13
Anurag Rana: workflow in those days.
6:14
Kris Rasmussen: Yeah, I think like kind of pre-AI, um the the kind of core value proposition, the Figma, the beachhead, if you will, was enabling designers to go from idea to more concrete artifact faster. And then using our creative uh canvas in order to align their team on the feature direction of their products. So they might mock up some screens, uh build a
6:34
Kris Rasmussen: design system to sort of align their company at large around how their their brand should be deployed consistently and appropriately, um, and then work through some key workflows and explore the solution space alongside other designers and engineers working with them on those features or products. Um and so to that end. uh, one of the things that's always excited me about working at Figma is it sort of combines
6:54
Kris Rasmussen: the sort of visual graphical nature of uh design with this sort of more uh pragmatic and important business narrative around helping your company align on where you want to go. And uh that's why we became such a vibrant place for entire teams, not just designers, to collaborate around these artifacts in our canvas. Um, and we started expanding our
7:15
Kris Rasmussen: product suite in turn. So we introduced Fig Jam during the pandemic as a way to bring a more approachable canvas to those collaborators. And we started to see, you know, other teams use Figma bottoms up, like engineering teams using Figma to to sort of brainstorm architecture diagrams and new systems and things like that. And also continue to work alongside designers. And then we
7:34
Kris Rasmussen: introduced slides as a way to kind of further refine um that alignment narrative. and help teams you know present their ideas to each other um and bring them to life and make them a little more crisp and and you know we continued from there. And so that's generally how people have historically used Figma. And I think that actually the opportunity going forward is is keeping them using it that way, but do it alongside
7:54
Kris Rasmussen: agents. It turns out that all this stuff that we do to align and work collaboratively and think, you know, as a team is also exactly what we need to be doing in order to align our agents and think alongside our agents. And so the workflow is changing in that we're seeing our customers not just build these collaborative artifacts to kind of you know hand off to an engineering
8:14
Kris Rasmussen: team or another stakeholder, but actually give them directly to agents using our MCP server or our new features like code layers and actually start to implement the real thing.
8:22
Anurag Rana: Yeah. So if I'm a banking, you know, if if if I work at a bank and I have Figma taking care of my digital application, now I'm assuming you get feedback from customers and then, you know, you go to the next iteration of that application. So, you know, my banking application now has very, you know, sophisticated an AI agent that's helping me out. So
8:42
Anurag Rana: I'm assuming is is that your biggest growth driver is as people adding more features to the application or is it developing of brand new applications?
8:50
Unidentified speaker (UNKNOWN): Yeah,
8:51
Kris Rasmussen: I think so I think that um it's a combination of both. So there's companies that are trying to expand into adjacent use cases and workflows. And there's companies that need to radically reimagine their product suite in order to capitalize on all the new affordances that this technology creates and the new expectations that come with it from users. So I think we're seeing uh expansion on
9:10
Kris Rasmussen: both fronts. And we uh just like every other company out there are also navigating this transformation ourselves and we're applying the best practices that we learn ourselves um back into our product suite to help others.
9:21
Unidentified speaker (UNKNOWN): Yeah,
9:22
Anurag Rana: no fair point. Chris, when we look at the SaaS industry in general, you know, I've always thought about a SaaS company with multiple products. So let's let's say Microsoft, for example, you know, I I use Excel separately, then I use Word separately, and I use Power World you know, PowerPoint separately. Now each application has its own interface functionality,
9:42
Anurag Rana: product identity. And then a lot of these applications became web applications. So you can collaborate with
9:49
Anurag Rana: your teammates. But now what it seems that your main interface is becoming an agent, and In that agent would help me design all of these things together. So it doesn't matter. I don't have to, oh yeah. well, you know, in theory open Excel or PowerPoint a Word anymore. I'll just go back to Copilot and say this is what I'm trying to
10:09
Anurag Rana: do. And everything gets done on the back end. Is that how you see the world evolving as well? Or do you have a different frame uh in mind?
10:17
Kris Rasmussen: Yeah, I think um there's there's a similar theme, but I'd I'd frame it a little bit differently in that uh I don't think the main interface is an agent per se. I think the agent is the glue. that connects all of these different interfaces. And so whereas previously without an agent, without this ability to kind of solve these ambiguous, non-deterministic problems, we really had to
10:37
Kris Rasmussen: encode the best practices into the workflows of these individual applications and teach the users to discover and use those applications at the right time. Now we have this uh incredible superpower to connect those different interfaces and input inbound input modalities and ways of working under a common interface that sort of democratizes the tools, which is that that sort of uh
10:56
Kris Rasmussen: agentic interface that we're all starting to become familiar with. But at the same time, it's not like everything I do in a spreadsheet should be done through uh an agentic interface. Like sometimes I really do just want to go click a filter and sort and poke around. And I don't actually want to spin up a new spreadsheet from scratch. I want to you know leverage the muscle memory and for affordances to. that I've developed with my,
11:16
Kris Rasmussen: you know, my own company's uh spreadsheet software that I've been using for a long time. And so I think that um although like chat as an interface and prompting is a really important new input modality, I think all these other input modalities are still relevant. And the best software is going to combine them uh into a cohesive whole so that it can kind of uh bend to the
11:36
Kris Rasmussen: workflow of the users, but also lean into the domain X. expertise and specialization that uh each of these individual companies have uh developed over the years.
11:44
Anurag Rana: So you still feel that the interface of the traditional software or the um would would would play a very critical role down the road?
11:54
Kris Rasmussen: Yeah, no doubt. I I think you see this in the in the evolution of the lab software themselves, right? They're building these richer desktop applications and introducing more. traditional UI affordances inside of their own agentic chat interfaces. Um, there is like a different in expectation in that the user doesn't necessarily want to learn as rigid of a workflow as
12:14
Kris Rasmussen: they might have had in the past, but they still want to have different ways of viewing information more richly and interacting with it rather than having to go back to the chat box. Um, I think a good example of this is is just writing in general, right? Like if you uh you if get, you know, you write a first draft and you ask AI to go rewrite it for you and make it better, it becomes pretty tedious every time you
12:34
Kris Rasmussen: see the AI kind of make a mistake or do something that's not really in your voice or consistent with your own expertise to go back to that prompt box and make all those individual changes. It's often a lot more direct and efficient to just grab the uh the document editor yourself and and start to kind of, you know, make those edits just like we always have. Um And I think the same thing is true of uh coding and and also
12:53
Kris Rasmussen: visual design as well.
12:55
Unidentified speaker (speaker_2): That's very interesting. Thanks for that. I mean um I remember I think Dylan in
12:59
Anurag Rana: one of
12:59
Unidentified speaker (speaker_2): the earnings call I can't remember it was six months ago or nine months ago when he started you know the conference call with going into Claude and start typing there. And you know I was a bit surprised. I said wow this is and and I think we were one of the first companies to do that. So you know, take me to a journey today. If I am a
13:19
Unidentified speaker (speaker_2): um designer, how do I go from um, you know, an AI tool to your software package to the point of shipping code out? So please would love to and any examples you could share, that would be really good.
13:32
Unidentified speaker (speaker_3): Yeah, so um we're we are we're trying to be flexible
13:35
Kris Rasmussen: in terms of workflow because great ideas can start anywhere. And so um there's sort of two different entry points. You could either start inside of our platform on our canvas using agent. So for example, um you might have already assembled a design system uh and and iterated to make it more agent friendly. And so rather than starting from scratch, you're gonna start with your design
13:55
Kris Rasmussen: system and prompt our agent to to build a new idea or workflow. By default, our agent will go and create uh sort of static anchor points like mocks and screens that sort of outline the user journey that you've given to it or it's reasoned about itself. And then you're gonna go back and forth with that agent and with a lot of our traditional
14:14
Kris Rasmussen: direct manipulation tools to refine that design and really make it different than the the default distribution that the agent or model might output by default. At the same time, you could equally start uh outside of our platform. inside of a coding agent or something like that. And then you can use our MCP server to bring that workflow inside of Figma. Maybe you you sort of
14:34
Kris Rasmussen: asked the model to to create a um a new uh like banking workflow or something like that um in code. And now you want to expand the state of that application to more easily visualize it and understand what the model was reasoning about when it when it built that prototype. This is one of the things that I think a lot of people don't understand is that by default, when
14:54
Kris Rasmussen: you prompt an agent, if you're not incredibly specific, you're gonna get sort of like the the default safe distribution of that model. It's gonna give you a workflow that's sort of the average of all things because that's that's effectively how they're trained, right? They want to be kind of safe by default. And it's not always uh explicit that every single thing that the agent did is exactly what
15:15
Kris Rasmussen: you wanted, right? A lot of it is accidental. For example, if you're a designer and you're you're implementing like a um a new uh banking workflow or something like that or a banking feature, you might have been very intentional about how you guided the model in terms of the user experience, but very unintentional in terms of how you guided it around data safety, data security, what APIs to
15:35
Kris Rasmussen: use, that sort of stuff. And so um it's not always the case that everything you create with an agent goes straight to production. Oftentimes you're building to think more so than building to ship. And there still needs to be this explicit curation where you need to help the other people on your team or maybe the agent that's actually going to implement it for real in your code base
15:54
Kris Rasmussen: understand what's next. And um, and that's a that's a really important workflow that we still see our customers adopting. It's something we see people do on our platform. And so recently we added new features to our Creative Canvas like specs and code to help people evolve through this journey. So they can start with prompting the agent to build a fully interactive prototype or some static mocks.
16:14
Kris Rasmussen: Um and then they can sort of evolve and ask the agent to expand upon what it was trying to do and create a a sort of handoff artifact either for a human in the traditional sense or for the next agent. And they can further iterate and curate that artifact with engineers, with product managers to really refine um the taste and judgment that they're trying to
16:34
Kris Rasmussen: imbue into the the the next agent and then use that artifact uh as context, if you will, um, to hand off to the next person on their team or or a coding agent on our platform or off our platform to go implement it for real.
16:47
Unidentified speaker (UNKNOWN): So
16:47
Anurag Rana: Chris,
16:47
Unidentified speaker (speaker_2): let me
16:47
Unidentified speaker (UNKNOWN): d
16:48
Unidentified speaker (speaker_2): dive a little bit into what you just said about if I am asking the generic agent about a banking workflow, and it gives me an answer that is a um, I guess the you said statistically, you know, uh an average of everything. But I'm looking for something extremely specific because I have a very specific need. I'm a bank with a
17:07
Unidentified speaker (speaker_2): different geography, different language, different regulatory needs. Um, then I have to go back and redo all that stuff. in your agent. Like how is that just ca increasing the total cost of ownership to uh the end customer or would they rather not start with your agent and just get it done over there?
17:26
Kris Rasmussen: Yeah. So we we have made sure that our agent starts with the right context by default. So for example, if your company has already built up a large repository of designs and design systems, it's possible to start with those those values that context already available to our agent. So it does the right thing more often than not by default relative to the context that you've already curated for it. But
17:45
Kris Rasmussen: even then, you're gonna want to iterate and refine the result. And so rather than starting from scratch, what you're actually doing is you're continuing to prompt the agent or work alongside the agent with the context that you've already produced and it's already produced to further refine the output. So every step along the way, you're building up more information. and more of a history of what you want and what
18:05
Kris Rasmussen: your intentionality, taste, and judgment is. And so each step gets you closer and closer to that final result rather than kind of throwing it all away and starting from scratch.
18:14
Unidentified speaker (speaker_2): That's a very good point. And you know, we have had uh multiple conversations with uh, you know, some of the larger SaaS companies, whether it's SAP or Workday. And one of the things we have heard is the importance of context. and the importance of workflows that are already embedded in the system. So tell me a little bit, if I am
18:35
Unidentified speaker (speaker_2): going to the application through an external agent, what are the things that are not available for me that is available onto the Figma or the, you know, your own agent uh that just gives me a better um output or a better result statistically?
18:51
Unidentified speaker (UNKNOWN): Yeah.
18:52
Kris Rasmussen: I mean, so the the first thing is that a lot of these uh these coding agents off our platform, they really work in these kind of linear siloed ways, right? They're not collaborative by default. Um, they're they're trying to help a user kind of progress through a linear series of prompts to a final result. Um what's different about our platform is that with the Canvas, we're
19:12
Kris Rasmussen: inviting the agents to actually be more expansive. in their thinking and and also like work alongside users to really explore the solution space more fully. So for example, uh if you bring your code onto our platform or you start with code on our platform, rather than working on a single code base at a time, you can more easily work with dozens of different code bases. And rather than focus
19:32
Kris Rasmussen: on the coding results, you can actually focus on the visual results that that the code produces. Um And so we we really believe. that as code becomes more and more commoditized and easier to produce, the value is moving up the stack to what you would call design. And we think it's really important that we help facilitate workflows that are much more divergent and
19:51
Kris Rasmussen: exploratory rather than just sort of these linear single player experiences. And so the coding agents on our platform and the agent at large is designed to understand Figma by default to get to better results with your existing design systems faster, and to do so in a way that facilitates and invites much more uh alignment and
20:11
Kris Rasmussen: collaboration from the individual user or the entire team that's working alongside these agents.
20:16
Unidentified speaker (speaker_2): That's a fair point. See, one of the ways uh the questions we get from investors is
20:20
Unidentified speaker (UNKNOWN): you
20:20
Unidentified speaker (speaker_2): know, if you are a user, who's
20:23
Unidentified speaker (UNKNOWN): you
20:23
Unidentified speaker (speaker_2): know, who's spending a lot of time onto your
20:27
Unidentified speaker (speaker_2): um, you could say, your own platform versus accessing a common, uh, different interface. So let me like let me throw it out and say if you're a large enterprise, you probably have your own agent that is you're designing to help your employees. You know, that can be connected on the back end to an HR system, to a finance system, and then like you know,
20:47
Unidentified speaker (speaker_2): let's say Figma as well. You also have companies. that run on Teams and Slack. So that's another way you are accessing the you know earth system in the back end. But now the mind share goes to those front end and not so much to the, you know, historically SaaS company. So how how do you balance those two
21:07
Unidentified speaker (speaker_2): or what what would you tell those people who are extremely worried that not every SaaS company will have that front end um, you know, mind share that you used to have.
21:19
Kris Rasmussen: Yeah. Well, so I think like at least in in Figma's case, I think that we still have that front-end mind share in the sense that we earn the right to build the best workflows and the best tools for a given set of users. And so for us, the the creative canvas is really a unique workflow that a lot of applications don't either have or don't put front and center and haven't invested in those same ways that we have.
21:39
Kris Rasmussen: And so we still s have um, you know, all these active users that are working directly inside of our product every single day. At the same time, we've always had integrations and APIs into these other uh platforms and tools that you just alluded to. And we think that's an important part of actually building a more complete workflow for our customers. Um we don't necessarily care
21:59
Kris Rasmussen: whether or not they started start inside of our canvas or outside of it as long as we are uh actually contributing value into that ecosystem and into those workflows. And inevitably we earn the right to to bring them back in in order to to facilitate this collaborative place to help them align with their team and with their agents. And I think the same stints true for many other types of software. I
22:19
Kris Rasmussen: think that um, you know, we've we've we've been in this world where APIs are a very important part of business software for a long time. And uh now those APIs uh have much richer and more dynamic capabilities thanks to agents. At the same time, um, we're still building these capabilities for humans and agents in a similar way that we
22:39
Kris Rasmussen: were for just humans. And I think that companies have a right to kind of um to show that they they actually have the domain expertise and the understanding to to build the best tools in this ecosystem. Aaron Beautiful.
22:50
Unidentified speaker (UNKNOWN): Powell
22:50
Unidentified speaker (speaker_2): You know, for your customers that have started to go this route of um
22:55
Unidentified speaker (UNKNOWN): you
22:55
Unidentified speaker (speaker_2): know AI adoption through agents, what are some of the
22:59
Unidentified speaker (UNKNOWN): you
22:59
Unidentified speaker (speaker_2): could say bottlenecks that they are struggling with most?
23:02
Kris Rasmussen: Yeah, so uh it kind of goes back to a point I was touching on earlier. I think that um in the early days, uh as a I say early days, I think it was like really the beginning of this year when a lot of these models got so powerful that it fundamentally changed the way people thought about. um running their teams and building software. Um so in the early days, I think everyone sort of assumed that now that you can kind of build a prototype from a
23:22
Kris Rasmussen: prompt, um, there there isn't really any need for collaboration. And you know, everyone's just gonna kind of converge into a single maker role. Um and I think there's some truth to that. I think there is definitely a convergence of roles and we're seeing more and more designers ship directly to production or more and more engineers participate in the design process.
23:39
Unidentified speaker (UNKNOWN): But
23:39
Kris Rasmussen: at the same time, like I said before, just because you can prompt a model to get a working prototype doesn't mean that every single decision you that the model made in the course of building that prototype is actually a decision you would want to ship to production. And say, you know, for example, if I uh if I prompt a model, like I said earlier, to make a change to a banking application or
23:59
Kris Rasmussen: prototype a new banking application, it's not necessarily going to gonna understand the nuanced kind of uh context history and and domain expertise that my banking engineering team has had to develop over decades in order to do so intentionally and thoughtfully and safely and securely. And so a lot of that code is is accidental, if you will, because the
24:19
Kris Rasmussen: the designer prompting it isn't really sweating the details of all those downstream systems. They're thinking first and foremost about the actual user experience. And that's sort of the domain expertise and the taste and judgment that they're imbuing into that product. And so while they might ship a PR to change the front end of the application directly to production, they're probably not going to take responsibility for
24:39
Kris Rasmussen: the the change to the kind of backing data systems because they're not the one on call who has to really sweat the details of that decision and and make sure that it's it's uh you know something that can be operated at scale robustly and securely. So I think this is this is one of the uh the insights that a lot of people are coming to. And um this is why we still see fit people using Figma
24:59
Kris Rasmussen: actually, even if they're working elsewhere, like with a coding agent, they still come back to Figma as a form of alignment and context creation before handing it off to the next team or the next agent.
25:07
Unidentified speaker (speaker_2): Yeah, that was what I've been was going to be my next question as to what behavioral change have you seen? Do you see people coming back to Figma to do one particular kind of task, or is it just a little bit of everything when they start uh, you know, working through an agent?
25:23
Kris Rasmussen: Yeah, I mean in many ways we still see the same behaviors that we've always seen in the sense that people are still using Figma as a place to more rapidly align um and curate the context that they then hand off to the the next part of their process. The big change is just that the next part of their process isn't always like handing off directly to an engineer. Sometimes it's handing off more
25:43
Kris Rasmussen: directly to an agent. Um, the other big change that we're seeing is that um, you know, we we are the place that a lot of companies curate and manage their design systems. And historically, I think those design systems were predominantly used by designers to then assemble new screens and workflows that were then handed off to their engineers. And now we're
26:03
Kris Rasmussen: actually seeing a lot of our customers assemble those design systems in such a way that uh imbues more uh context so that they can be more effectively used by agents. And so they'd still have control over the consistency and quality of their application, even as they scale out the development through agents.
26:18
Unidentified speaker (speaker_2): Chris, one of the arguments I've made over the last 12 to 18 months is some of these SaaS vendors, especially those that are critical systems, are of record, people go to them because of trust and because of their history of maintaining their core systems. Now, if something goes wrong with that design that was made on the banking
26:38
Unidentified speaker (speaker_2): app, um, you know, how do you make sure you tell the client that listen, this is not my fault. This is something that the agent, like, how do you protect your reputation and you know, your um intellectual property just because the um an agent made something or or you know, an AI agent made something that really screwed things up?
26:58
Unidentified speaker (UNKNOWN): Yeah,
26:59
Kris Rasmussen: I mean, I think this is a this is a a very big challenge with like uh agentic coding tools and agentic design tools in general. And the way we solve for it is we try to come up with ways of helping the user and the agent do things more explicitly and make some of these decisions that might be implicit. more explicit. Um and so, you know, for example, helping the
27:19
Kris Rasmussen: user expand a uh a given kind of prompt that might have produced a single kind of banking flow for a company into the actual individual screen so that at a glance they can understand it and take more responsibility over what was produced and then curate the results and, you know, change things as needed. Uh, I think it's really important that.
27:39
Kris Rasmussen: we we design these sort of new agentic interfaces so that they're more transparent with what they're doing um and they make their own design decisions more explicit versus implicit and hidden behind the scenes. Um this is a key part of the the design workflow and always has been. Like how do you actually help your stakeholders understand what you have in your
27:59
Kris Rasmussen: head by expanding and exploding it into a uh a series of um key information and states that that really helped you curate at the right altitude and and make the right decisions.
28:10
Unidentified speaker (speaker_2): Another thing I remember from your last earnings call, there was a lot of discussion around helping designers or coders to build a product or prototype first, see how it is, rather than wait till the final moment of uh, you know, it's like so g give us a little bit of background of what's kind of the thesis behind it.
28:28
Kris Rasmussen: Yeah, I mean the the beauty of these models and sort of the commoditization of code in general is that you can get further faster, uh much faster than you used to be able to in a in a more traditional kind of, you know, pure human in the loop workflow. And so as a result, um whereas you know in the past a uh a designer might have started uh
28:48
Kris Rasmussen: the exploration um but then still relied on an engineering team to go and implement it and turn it into a richer prototype. Uh now what we see is that there's affordances within our platform for designers to actually go all the way there and to build that fully interactive prototype um alongside all of the the additional kind of uh context and you know,
29:08
Kris Rasmussen: curated information that they want other people to really understand at a glance. Um, and this not only helps them drive stronger. alignment and get to better results faster, it also helps them think faster, which means that they spend more time thinking about what the experience should be, to your point, rather than um just uh just, you know,
29:28
Kris Rasmussen: get doing just enough to to hand it off to the next step and then waiting until someone else implements it and then think again. So it shortens a lot of these thinking loops and helps people build better experiences.
29:37
Unidentified speaker (speaker_2): And um, you know, just going into the infrastructure world, just because you have some background in it, you tell us a little bit about the models that you're using, the infrastructure that you use, and has that changed in the last few years with everything that's happening with agents?
29:52
Kris Rasmussen: Yeah, we've positioned our infrastructure in order to take advantage of whatever models are best at any given task. Um so I think what's changed is that the the number of models that are really good is is quite wide and expansive now. And uh we've also changed our infrastructure to better facilitate training loops for some of our own models so that
30:12
Kris Rasmussen: we can help users accomplish the same results um with equivalent or better quality, uh faster and cheaper. And so um the biggest shift is probably that we've been directing more and more of our internal traffic to our first party models. for specific subtasks of these agentic workflows inside of our our Figma agent product, which just went to GA.
30:32
Unidentified speaker (speaker_2): Wow, that's very interesting, Chris. We've had actually the pleasure of having, you know, co-founders of uh um, you know, Serval and Sierra and Finns on the podcast. And I was I was very intrigued that how many of those companies are actually using multiple models, open models and their own first party models.
30:52
Unidentified speaker (speaker_2): So I'm I'm very curious as too, because you know, one would expect that your your own models would require a lot of capital and a lot of uh research to come up with. You know, perhaps give us a little bit of a background on your own models and you know how it came about and what kind of structure you're using.
31:10
Kris Rasmussen: Yeah, so um I think that the the first thing is we're not building our own models from scratch as
31:14
Unidentified speaker (UNKNOWN): well.
31:15
Kris Rasmussen: a lot of the the labs historically have had to. Um we're benefiting uh on the backs of a lot of these open weight models that have merged in recent years and and really picked up spe steam more recently. And so um we're also not building our models right now to be just these sort of general purpose jack of all trades models. We're optimizing them for the workflows and trajectories
31:35
Kris Rasmussen: that we see our users doing. So we get to focus uh a lot more than I think, you know, a lab might be able to focus. And as a result, we're able to do these things, uh, you know, train these models with West capital uh more efficiently.
31:48
Unidentified speaker (speaker_2): And do you use a harness to have multiple models in the background or do you pick or you know, uh you know, we'd love to learn a little bit about that as well.
31:56
Kris Rasmussen: Yeah, yeah. As as I mentioned before, our our infrastructure is designed to sort of take advantage of the best models for any given task. Um we don't ever want to be in a place where a new model comes out that's better at one thing and and not be able to help our users adopt that model. Um and so while we do uh with our own agent product uh make decisions on behalf of users in terms of how to optimize for
32:16
Kris Rasmussen: quality performance and cost, we've also positioned our our infrastructure with uh a harness backing that allows us to kind of pick and choose different models for different tasks. So for example, you might use a frontier model in the Figma Make product to generate an interactive prototype, but under the hood, we will use different image models to generate the uh the raster
32:36
Kris Rasmussen: imagery within it rather than being wetted to a particular provider. And we think this is a really important part of our strategy, being model agnostic so that we can really help our customers understand which models are best for any given task and build the best tool for their workflows.
32:49
Unidentified speaker (UNKNOWN): And
32:49
Unidentified speaker (speaker_2): is it driven by a cost? concept or is it driven by usability or functionality?
32:55
Kris Rasmussen: I think first and foremost, we're always thinking about quality. That's core to our brand. We always want to build the best experience for our users. And then from there, there's uh there's a lot of like obvious ways to further optimize for speed and cost as we start to see how these models are used and what they're good at. Um And as I mentioned, we also have this other lever now where we can train our own models off of.
33:15
Kris Rasmussen: the backs of these open weight models to further uh improve these trade-offs.
33:19
Unidentified speaker (speaker_2): So and you know, what is the benefit of the first party model, as you said, uh using a you know existing open uh model or open weight model versus a foundation model?
33:30
Kris Rasmussen: Yeah, I mean the the benefit is that we have like vast quantities of design data and trajectories from our customers. And so we're able to fine-tune these models. to not just be better at design, but also to be better at Figma. And this is something that the labs don't have access to. They don't have the the data that we have. So we can look at the trajectories of how agent is
33:50
Kris Rasmussen: now performing in the wild and we can see places where the the open weight model by default might be inefficient. It might take a wrong The turn and have to kind of back up and try again. Um, And then we can use these trajectories to further fine fine-tune tune these models. to bring efficiencies both for our business and for the customers.
34:06
Unidentified speaker (speaker_2): So, Chris, when we have looked at the entire SaaS industry and we get a lot of questions on it, the number one area people show concern is about uh companies that sell software tied to creativity, to design. And, you know,
34:21
Unidentified speaker (speaker_2): I'm sure you've seen, you know, how Adobe has done in the last couple of years. And so tell me. a little bit about why you think if these frontier models really could become even better in the next few years, why wouldn't they be able to do the entire life cycle of what, for example, Figma is currently
34:41
Unidentified speaker (speaker_2): offering to customers?
34:42
Kris Rasmussen: Yeah. So first off, I think it's it's really important to remember that one of the hardest problems with these models is aligning them to what humans want. And the reason that's such an important problem is like we want to be in control of these models. We have our own unique taste and judgment that we want to imbue in the output of these models. And that's a really interesting and fundamental challenge for everyone. The labs
35:02
Kris Rasmussen: alike. And to me, that's like the essence of design. And I think it's something that we've we've been doing for over a decade really, really well. And companies trust us to be the place that they curate the context and they get the economy. economies to scale in order to really ensure that they're doing this effectively and efficiently and that they're doing so in conjunction with
35:22
Kris Rasmussen: the alignment of their team and their and their business as well. I think that the opposite question is equally relevant, which is why why are the labs better positioned to do these things, right? Like why are they better positioned to understand all this domain expertise and context just because their models are really good doesn't mean that they're the best people to build the
35:42
Kris Rasmussen: tooling and the workflow for humans around it. I think that a lot of these models have uh become increasingly commoditized across the different uh labs and you know, open weight models alike. And so it's it's not just the labs anymore that have access to this technology to build this next generation of software. I think it's available to everyone and it's available to us at Sigma too.
36:01
Unidentified speaker (speaker_2): You talked about token. tokens. Um, you know, how do you manage them internally? And do you manage this, the tokens on behalf of your customers or they have to do it themselves?
36:12
Kris Rasmussen: Yeah, so it internally, uh in terms of like our own internal development, um, we uh we haven't deployed hard caps, um, although we're starting to explore ways of introducing soft caps to help people be more cognizant of what is and is not consuming a lot of tokens.
36:27
Kris Rasmussen: We're really trying to foster a culture internally where there's a lot of experimentation and learning. And so what we do, and we've seen a lot of our customers do this too, is we we highlight the best use cases at our week, our bi-weekly all hands in terms of helping people understand how design and development is changing and how to use these models most effectively. Um and then uh yeah, for our
36:47
Kris Rasmussen: customers, we give them controls in terms of spend.
36:51
Kris Rasmussen: and the ability to procure a fixed set of uh tokens or you know, credits, if you will, um, for each seat. Um and we also embed some of these credits inside of the seats themselves.
37:03
Unidentified speaker (speaker_2): And if the customers are looking for more tokens, I mean, is there do you recommend them how to do go you through go about it? Or you know, it's just that you know, you turn on the um uh spigot and get them more tokens. I'm I'm very curious as to you know down the road do will companies manage that themselves or would the software vendors help
37:23
Unidentified speaker (speaker_2): them in that journey?
37:24
Kris Rasmussen: Yeah we're we're trying to help them in the sense that we we know that our agent itself needs to be as efficient and effective as possible. And so whenever we're doing our own hill climbing um on our own item agentic to capabilities, we're always looking at cost, which is oftentimes correlated with latency in conjunction with quality. Um, and and we think of needing to always get to the
37:44
Kris Rasmussen: same or better result uh with with cheaper cost and latency.
37:47
Unidentified speaker (speaker_2): Yeah, I'm just curious as to how whether companies themselves will be in the business of managing tokens for multiple of their products. Because I've never, you know, you know, one could uh JP Morgan could argue that listen, I am better at getting you know better prices on tokens from these. model companies. Let me manage that on my end.
38:05
Kris Rasmussen: Yeah, that's uh it's an interesting angle. And uh we do have like an MCP server, so people can still use our software and capabilities with third-party uh agents as well. So we we do give customers some flexibility and choice. Um but ultimately I think that this is a very complex evolving landscape. And I think that within a given domain, uh certain companies have the
38:25
Kris Rasmussen: incentive and the motivation to to figure out how to help companies do this effectively, how to find the right mix of models to get to uh equivalent or better results with uh improved latency and cost. So I think that uh it's it's evolving as we go and um we give our customers control in the sense that they have the ability to um purchase additional
38:45
Kris Rasmussen: add-on add on credits um or turn on pay as you go if they if if they want to so they still have control.
38:50
Unidentified speaker (UNKNOWN): And
38:50
Unidentified speaker (speaker_2): you know when you think of the SaaS industry in in general, it was primarily a seat based subscription model as so many new things coming out right now, outcome based pricing and usage and other things. How do you think the world evolves, you know, a few years from now?
39:06
Kris Rasmussen: Yeah, it's uh it's gonna be hard to predict exactly how it evolves. Um I think that customers have shown an appetite and if not an eagerness to really um pay for outcomes. Um and so I think that we're seeing sort of these modern business models evolve to have seat based components and also consumption based components as a natural
39:26
Kris Rasmussen: evolution of that. I I can't predict exactly where it's all going, but I I think that um customers have already shown that they um they're willing to pay for better results um as long as it's competitive.
39:37
Unidentified speaker (UNKNOWN): You know,
39:38
Unidentified speaker (speaker_2): I'm assuming only a small percentage of your user base currently use uh external agents to design. You know, how do you see that mix evolving in the next three years?
39:49
Kris Rasmussen: Um so actually it's it's a pretty uh sizable percentage of our user base that both uses external and and Figma agents to design. Um we see a mix of both. For example, there's customers who predominantly use the Figma agent to design inside. of our platform. And they might even, you know, use some new features like code layers to turn those into working interactive
40:08
Kris Rasmussen: prototypes and better handoff artifacts. And then they connect their MCP server to a third-party coding agent and they use that to then go and implement it inside of their production code base. So we're we're already seeing a pretty healthy mix of both workflows, depending upon the use case. And I think we're gonna going to continue to see this
40:28
Kris Rasmussen: evolve in the sense that more and more sort of round-tripping between these two different workflows, which is a core part of our product strategy, you know, going from code to design, third-party agent to in-house Figma agent on the canvas, and just really making sure that our business is well positioned to help provide the right tools, um, models,
40:48
Kris Rasmussen: and and uh flexibility for the customers. to navigate that.
40:52
Unidentified speaker (speaker_2): Got it, Chris. Chris, we are almost running out of time. Perhaps my final question for you is you know, where what do you think is the biggest opportunity for Figma looking ahead?
41:02
Kris Rasmussen: Um I think that there's this this huge opportunity to really help customers understand how to navigate the changing expectations of software and do so in a way that still gives them explicit control over the the taste and judgment of the uh the results that they're
41:21
Kris Rasmussen: producing. So I think that we have this this tremendous opportunity to help customers work more iteratively, collaboratively to really embrace the role blending that is now possible with these agents, but do so in a way that still ensures that they are in control um and that they are able to align with their team alongside agents to
41:41
Kris Rasmussen: build better results for their customers.
41:44
Anurag Rana: Great, Chris. Chris, thank you
41:45
Unidentified speaker (UNKNOWN): so
41:45
Anurag Rana: much for your time today.
41:47
Unidentified speaker (UNKNOWN): Thank
41:47
Kris Rasmussen: you. It was great chatting with you.
41:48
Unidentified speaker (speaker_2): For our listeners, we want to thank you for tuning in. If you like the episode, please subscribe and leave a review and check back to hear conversations with the leading disruptors in the tech landscape.
41:59
Unidentified speaker (UNKNOWN): If
41:59
Unidentified speaker (speaker_2): you want to learn more about our research, including deep dives on topics like AI, check our work on the Bloomberg Terminal at BI Co. We also want to thank Aditya Samani for his help in editing this podcast. This is your host, Anragrana, signing off.
This page was produced automatically: the audio was transcribed with nvidia/parakeet-tdt-0.6b-v3+sortformer-v2.1 and interpreted by Qwen3.6-35B-A3B-UD-IQ3_S.gguf. Transcription and interpretation can contain errors; always confirm against the original episode. Nothing here is investment advice.

More from Tech Disruptors

Tech Disruptors

Honeywell CTO on AI and Physical-World Autonomy

Episode discussion: Honeywell CTO outlines AI-driven autonomy strategy, citing edge computing and deterministic guardrails for industrial applications

Speakers’ claims: Mustafa Okur interviews Suresh Venkatarayalu, CTO and President of Forge at Honeywell, regarding the company's pivot from automation to autonomy. Venkatarayalu explains that Honeywell has simplified its portfolio to focus on commercial buildings and industrial process automation, leveraging its Forge platform to integrate AI with deterministic control systems. He describes a three-phase AI strategy involving internal knowledge models, partnerships with Nvidia for open-source models in secure hardware, and hybrid architectures for data sovereignty. The discussion highlights use cases in reducing false alarms in ethylene plants and building management systems, as well as collaborations with Google, LS Electric, and Dangote to expand ontologies and meet cybersecurity mandates.

Artificial IntelligenceIndustrial AutomationCybersecurity
52 min3 assets18 claims to verify
Tech Disruptors

IBM CEO on Enterprise AI and Software Strategy

Episode discussion: IBM CEO Krishna outlines enterprise AI adoption stages, open-weight model economics, and software consolidation trends.

Speakers’ claims: IBM CEO Arvind Krishna discusses the state of enterprise AI, noting that only 10-20% of companies have moved from experimentation to scaled deployment. He identifies coding, customer experience, and enterprise operations as key use cases, predicting that tech spend will double as a percentage of revenue over the next decade. Krishna argues that data accessibility matters more than cleanliness for AI agents and highlights IBM's internal success with AI automating HR and procurement processes. He forecasts that open-weight models will capture 80% of enterprise token usage within three years due to cost and control advantages, drawing parallels to the adoption of Linux.

Enterprise AISoftware StrategyOpen Source Models
48 min2 assets20 claims to verify
All insights