The MAD Podcast with Matt Turck

Episode: Inside Canva’s $3B ARR AI Design Rocketship — CTO Brendan Humphreys on Magic Studio & Canva Code
Guest: Brendan Humphreys, CTO of Canva
Host: Matt Turck
Air Date: June 20, 2025

Overview

This episode offers a deep dive into Canva’s remarkable journey from a startup to a $3B ARR, 230M MAU powerhouse, spotlighting its rapid integration and deployment of generative AI. Brendan Humphreys, CTO, shares insights into Canva’s product philosophy, tech stack, organizational growth, and lessons from scaling AI-first features to hundreds of millions of users. The conversation spans product innovation, the challenges and opportunities of AI, organizational scaling, engineering culture, and Canva’s broader mission-driven ambitions.

Key Discussion Points & Insights

Canva’s Growth Story and Engineering Philosophy

• Canva's Scale: Just surpassed $3B in ARR, growing 40–50% annually and profitably for seven years; 230M MAUs, 24M paying users, customers in 190 countries, used by 95% of the Fortune 500 (03:14–03:39).
• Engineering Team: 2,300 out of 5,000 employees are engineers, highlighting the company’s innovation pace (09:23–09:43).
• Organizational Journey: Brendan’s path from the 12th employee to CTO described as “a little surreal” and a succession of “fire drills” (02:18–02:33).
• Talent Selection: Emphasized the importance of founder quality, learning opportunities, and collaborative culture over “famous names” (05:22–06:16).
• Founder’s Influence: “Mel recognized that the entirety of the design process involved so many different tools... an opportunity to consolidate it into a platform that was web-based. That was the source of our success.” —Brendan (03:14–03:39)
• Engineering Culture: High agency, crossing swimlanes, and generalist mindset is key: “Everything’s your problem when you own a company.” (06:28–07:23)

Distributed Teams & Remote Work

• Hybrid Model: Originates in Sydney but has hubs across Australia, remote teams globally (10:01–11:15).
• Autonomy via Product Seams: Key to successful distributed engineering organization: “The key is to find ways for teams in incompatible time zones to be autonomous.” (11:41)
• Async Collaboration: Heavy utilization of Canva’s own doc suite, async comms, and minimization of cross-timezone meetings (13:02).

AI Strategy & Innovation

Early AI Investment & Platformization

• Long History with AI: In production before the generative wave; built ML team in 2017 for recommendations and propensity models (15:47–17:23).
• Third-party & Proprietary Blend: Aggressively integrates latest third-party models (OpenAI, Google, etc.) while developing in-house ML and foundation models via Leonardo acquisition (17:33–18:46, 42:15–43:24).
• Platform Abstraction: Major investments in platform abstraction allow swapping and experimenting with AI models rapidly (00:00–00:29; 18:46–19:10).

Product Features & AI Integration

• Product Velocity: Released Magic Studio, Dreamlab, Canva AI, Canva Code, Canva Sheets, etc., at “breakneck speed” (19:10–21:02).
• Magic Studio: Collection of AI-powered tools (text and visual)—automatic text, image editing, Magic Grab, Magic Arrays (19:44–21:02).
• Canva Code: New feature letting users build interactive widgets in designs; idea to 100M users in three months (00:00, 18:46–19:10).

Agentic Solutions & Copilots

• AI Agents: “We have behind Canva AI quite sophisticated orchestration that you could call an agentic solution… inferring intent… and executing on that action.” — Brendan (23:31–24:59).
• Agent Future: Noted the trend where Canva may operate both as a destination and as a composable tool within agent chains (24:59–26:21).

Democratizing AI Skills in Engineering

• AI Empowerment: Focused on democratizing knowledge—“DBA is not really a skill anymore… we like to think the same will happen with AI.” (26:43–28:23)
• Mindset Shift: “The non-determinism is a tricky thing… you have to move to testing with evals, which are much more probabilistic.” (29:10–29:50)

Generative Coding, Vibe Coding, and Developer Experience

• Generative Coding Tools: Omnipresent and embraced—~30% productivity gain in PR merges for engineers using AI tools (32:54).
• Caution on Vibe Coding: “We have a rule that you need to own the output of the tool. That means you have to understand it as if you’d written it yourself…” (30:48–32:25)
• Review Challenges: “It’s very easy to write a small prompt and end up with volumes of code. We’re seeing some success with AI assisted code review. Author with AI superpower will be matched by reviewer with AI superpower.” (33:29–34:35)
• Impact on Hiring: More productive engineers mean slower hiring, but focus on first principles, strong CS fundamentals and empathy for grads (34:40–36:59).

Empathy and Collaboration at Scale

• Empathy in Engineering: “It’s about getting in someone’s head and understanding... their point of view.” Brendan advocates for improv techniques (“yes, and...”) and steel-manning to foster healthy collaboration (37:04–39:21).
• Importance at Scale: Vital for productive distributed teams, especially with 2,300 engineers over continents.

Build vs. Buy, AI Stack, Security & Enterprise Push

• Pragmatic Excellence: “We are ruthlessly pragmatic… We want to leverage best-of-breed models… but put a platform abstraction over that so we can chop and change quickly.” (40:05)
• Stack Choices: Uses Vertex AI, Bedrock, proprietary model serving; eval with Weights & Biases plus internal framework (43:54–44:51).
• Open Source Model Usage: Llama, Stable Diffusion, Segment Anything, and more in production (43:24–43:44).
• Dealing with Hallucinations: Compositional heuristics keep things deterministic when required, especially for enterprise and brand requirements (45:18–46:06).
• Security: Up-skilled team—“very careful about prompt injection,” especially as enterprise focus grows (46:12–46:29).
• Enterprise Lessons: Invest early in scalable region abstractions, data residency, enterprise-level controls—hard to retrofit (47:10–48:26).

Technical Debt, Shipping Philosophy, and Product Ownership

• Technical Debt as a Tool: Willingness to run up “high interest credit card of technical debt” at key moments—“the important thing is to recognize it and make intentional decisions” (48:44–51:15).
• Iterative Shipping: Features behind feature flags and heavy dogfooding culture; feedback loop between product team and company-wide usage is critical (51:47–52:59).
• Product Ownership: “Everyone is a product owner, they really own the product, they care about the product, they think deeply about the product.” (51:47)

Vision and Mission

• Ultimate Ambition:
  • “Mel has a two-step plan: create the world’s most valuable company and then do the most good in the world.” (53:14)
  • “I think software engineering changes to a much more fun job… you're able to orchestrate software with these agents helping you and become much, much more productive.” (53:14–55:04)
• Upcoming Features: Integration of Google’s VO3 video generation model, more AI features announced for October and beyond (54:15–55:04).
• Purpose-Driven: Most founder equity pledged to Canva Foundation for global causes; $200M Creator Fund for contributions to AI models (55:30–56:12).

Notable Quotes & Memorable Moments

• On Platform Abstraction & Speed:

  “We invested enormously in putting a platform abstraction over a lot of the third party AI… allows us to rapidly switch out models and experiment. When new models come onto the market… we can get them in front of users… very, very quickly.”
  —Brendan (00:00)

• On Picking Winners:

  “I don’t think I have great advice on how to pick winners. Look, I’ve been very lucky... The thing that convinced me was the founders’ expansive ambition—and they’re just really nice people.”
  —Brendan (05:00–06:16)

• On Organization Scaling:

  “At early stage companies, you’ve got to be a generalist. That’s really powerful, and when you run your own company, everything's your problem.”
  —Brendan (06:28–07:23)

• On Product Shipping:

  “It certainly feels like breakneck speed at times. We’re very focused on building a set of composable components with powerful APIs, so product teams can move very, very quickly.”
  —Brendan (26:43)

• On Generative Coding:

  "You couldn’t pry these tools out of the hands of engineers... But you need to own the output of the tool."
  —Brendan (30:48–32:25)

• On Empathy:

  “Empathy... it’s about getting in someone's head and understanding... in a large engineering organization, it’s absolutely essential for productive collaboration.”
  —Brendan (37:04)

• On Technical Debt:

  “Technical debt is an extremely useful thing when you’re a startup... The important thing is to recognize it, make intentional decisions... Striving for pragmatic excellence.”
  —Brendan (48:44–51:15)

• Company Vision:

  “Mel has a two-step plan for Canva: create the world’s most valuable company and then do the most good in the world.”
  —Brendan (53:14)

Timestamps for Major Segments

| Timestamp | Topic | |-------------|---------------------------------------------------| | 00:00 | Intro to Canva’s AI platform abstraction | | 02:11 | Canva’s numbers and business growth | | 05:00 | Brendan’s motivation for joining Canva | | 10:21 | Distributed/hybrid team structure | | 13:02 | Async collaboration tooling | | 15:47 | Early AI efforts and Kaleido acquisition | | 18:46 | Productizing AI features at scale – Canva Code | | 19:44 | Walkthrough: Magic Studio, Dreamlab, Canva Code | | 23:31 | AI agents and orchestration in Canva | | 26:43 | Shipping AI features and engineering culture | | 30:48 | Generative coding tools, “Vibe coding” philosophy | | 32:54 | Productivity impact of AI on engineering | | 34:40 | Hiring and skillset for AI-era engineers | | 37:04 | Empathy and collaboration at scale | | 40:05 | Build vs. buy: AI stack decisions | | 45:18 | Handling hallucinations and guardrails | | 46:29 | Enterprise product requirements | | 48:44 | Technical debt and intentional choices | | 51:47 | Shipping philosophy and quality control | | 53:14 | Canva’s mission and roadmap | | 54:15 | Upcoming features: VO3 video model, AI roadmap | | 55:30 | Canva Foundation and the creator fund |

Conclusion

This episode offers a comprehensive window into how Canva balances explosive growth, relentless product innovation, and a mission-oriented ethos, all while adapting to and shaping the cutting edge of AI. Brendan Humphreys’ candid reflections provide a must-listen perspective for anyone building at scale in AI, engineering management, or product at the intersection of creativity, productivity, and technology.