Cameron Adams on Canva: Designing for Delight, Depth and Change

Canva co-founder and chief product officer Cameron Adams tells Harry Stebbings that speed matters, but a product must also reach a quality level that excites people. He credits users sharing a delightful experience with helping drive Canva’s growth.

Adams met Melanie Perkins and Cliff Obrecht through his former Google manager Lars Rasmussen, with whom he had worked on Google Wave. Perkins and Obrecht were running a school yearbook business and wanted advice on moving beyond Adobe Flash to browser-based design.

Their first meeting began with a misunderstanding: Perkins thought Adams was there for a developer interview, while he had come to advise on technology. Over coffee she described a bigger vision of making design accessible to people who had not previously had those tools. It appealed to his background in both graphic design and computer science.

Adams was then building another startup, Fluent, which expected a funding round that did not materialize. After the venture struggled, his wife suggested reconnecting with Perkins and Obrecht. They had not yet found the technical co-founder they needed, and the three joined forces in 2012.

Asked how much success comes from luck, Adams emphasizes preparation and exposure to opportunities: talking to people, learning technology and trying ideas. Many of his earlier projects did not work because of timing or team fit. He considers Google Wave an early collaboration idea whose browsers and users were not yet ready.

Investors urged Canva to launch quickly, reflecting the lean-startup thinking popular at the time. The founders resisted because they wanted a product of sufficient quality. Adams says that does not mean taking five years to launch; it means finding the point at which users are excited enough to tell others.

Small details can show users that a product was crafted with care. Adams mentions animation on Canva’s landing page and playful Easter eggs, including a duck that appears after a user uploads 100 images. Those moments matter because they make the experience interesting rather than merely functional.

The key onboarding moment is when someone creates something personal and realizes they can design. Adams says delight should occur during the first journey, although not necessarily in the first 30 seconds. The sign-up experience should set the tone and build toward that realization.

Canva aims for simplicity at the surface and depth underneath. Adams describes an 80/20 rule: a newcomer should access most of the value quickly, while continued use reveals more sophisticated capabilities. A user might only later discover typography and learn to become a better designer.

The move into enterprise required learning what large organizations actually need. Adams says an earlier Canva Enterprise launch was essentially a way to contact sales or sign up many users; it was not yet a full enterprise product. In hindsight, the company had not listened closely enough to that audience.

Asked about specialized AI tools unbundling design workflows, Adams argues that a platform joining those workflows has substantial value. He sees hype and churn among narrow tools, though this is the view of a Canva executive competing in the same market.

Stebbings worries that abundant AI-generated content will make discovery harder. Adams initially says strong content can rise, then agrees discovery and the creator’s full presentation matter. He uses Taylor Swift’s marketing, personality and tour strategy to illustrate how distribution complements the work itself.

Adams cannot identify a single largest adoption spike, but points to text-to-image as pivotal. Canva launched the early feature in about six weeks using Stable Diffusion, he says, reaching a product with around 100 million users. The launch started a broader period of AI experimentation.

He welcomes competition among model providers. In his view, better APIs and more standardized interfaces are making it easier to change providers, though he acknowledges that switching was harder in the early days. The speed and cost of switching can depend on the implementation.

Stebbings asks whether reliance on outside AI providers disrupts product roadmaps. Adams sees the rapid pace as typical of an early technology and predicts that innovation and interfaces will become more settled over the next five years. This is a forecast made at the time of the interview.

He expects interaction with AI to become more nuanced than a text box or voice prompt. Canva learned that a blank page can leave users unsure where to begin; an AI assistant should help people explore and iterate, not merely respond to a fully formed initial request.

Magic Media began as a text box, but many users cannot describe an image’s style precisely. Adams says prompting is a useful skill, yet hopes it will not become the defining measure of a product manager. If prompting remains the only interface in five years, he argues, product builders will have lacked imagination.

He compares a possible mature model market to cloud infrastructure: several major providers, some switching costs, and continued competition on features. This is his projection, not evidence that any particular AI market structure is inevitable.

At the time, Canva included Magic Media and other AI capabilities in Pro or Enterprise subscriptions. Adams says AI usage adds costs but had not materially hurt margins in his assessment, and that model costs were falling. He argues that adding value over time is part of sustaining a subscription.

He expects stand-alone AI add-on charges to become less important as capabilities become commoditized and included in basic products. Canva’s approach, he says, is to use AI to make a better product and continue its mission of broadening access to design. This is a commercial strategy and prediction, not a universal outcome.

AI has made product work more research-oriented, Adams says. With established web technology, teams could often look at a mockup and know how to build it. With a new model-dependent feature, engineering must first establish whether it can produce acceptable quality in real situations.

He gives enhanced voice as an example: a one-click tool to clean up audio in a video. During development, a team might find that a use case such as a noisy taxi does not work well, forcing product and design to adapt to technical constraints. He describes a closer dance between engineering, design and product.

Canva typically releases features to gradually larger cohorts, such as 1%, 5%, 10%, 20% and then everyone. This lets the company observe server load, quantitative and qualitative responses, and bugs. Depending on the feature, it may also segment tests by geography.

Geography is useful when payment habits differ, for example between cards, local providers and prepaid credits. Canva can evaluate how these options affect subscriptions and retention. The right test depends on what the team is changing.

A major interface change requires more care than an improvement to image quality. Adams calls the process “change boarding”: explain the benefit and help people move into the new experience, particularly when they may be in the middle of urgent work and expect familiar controls.

Canva gives some enthusiastic users and professional communities early access, but generally avoids flooding everyone with advance messages that may be ignored. For the interface update called Glow Up, users saw an explanation and could skip the change at a critical moment.

Adams estimates that fewer than 1% clicked skip, correcting a higher number suggested in the conversation. He says the option was designed for situations such as someone about to deliver a presentation. The transition itself is designed and user-tested before a broader release.

An early product mistake was trying to build a social network because investors were excited by Pinterest and Instagram. With a small engineering team, Canva assigned one engineer for nine months, but the experience did not fit the founders’ passion or their core product vision and performed poorly.

Many Canva users create private material and do not want comments or likes on it. Stebbings asks whether a larger commitment might have made the social feature work. Adams thinks it would have required a different product and says the lesson was to trust the company’s own convictions more.

Product reviews happen continuously rather than on a fixed weekly schedule, usually remotely. The cadence becomes more regular ahead of a specific milestone. Adams, Perkins and head of product Rob often attend alongside the relevant specialists, product managers and designers.

At the time of the interview, Adams says Canva had about 100 machine-learning engineers, including staff in Vienna from the Kaleido acquisition and in Sydney. Their work ranged from behind-the-scenes data tasks to visible features such as background removal, image generation and writing tools.

Resource allocation between foundations and new features is a continuing judgment rather than a fixed percentage. Adams recalls investing heavily in security several years earlier to build user trust. Not every important investment produces a visible new feature or immediate revenue.

He estimates Canva had roughly 4,500 employees, including about 1,700 engineers and more than 2,000 people across engineering, product and design. If all product and technology work is counted as research and development, it represents a large share; he does not provide a precise narrower R&D figure.

Although investors often compare Canva with Adobe, Adams says Canva set out to reach people who did not previously design. He prefers to organize the company around its own vision and customer impact rather than a common enemy. This is his positioning as Canva’s co-founder.

Canva staff use their own products, while also trying alternatives where useful. Adams is interested in newer marketing tools that bring creation, measurement and spending decisions into one workflow, rather than requiring a separate product for each step.

In the quick-fire round, Adams says he keeps relearning the value of doing something unusual. He recalls initially worrying that a rap video at Canva Create could be awful, but trusted Perkins’s conviction. The video attracted debate and, he says, tens of millions of views while communicating enterprise features.

He pushes back on advice to launch and iterate without a strong underlying vision. Building a company for decades requires belief in the problem, he argues, particularly through difficult periods. Asked which product job he would try, he chooses a music company such as Ableton or Serato because he loves DJing and music production.

Adams says he and his wife started Wedgetail to focus on what he calls the nature crisis. He distinguishes climate change from the wider health of forests, animals, water and ecosystems, while noting that the two are connected. These are his framing and priorities.

Stebbings raises a concern that environmental transitions in wealthy countries could harm jobs and livelihoods elsewhere. Adams argues that developed countries can have effects beyond their domestic emissions, including through exported resources, and that countries can pursue development in more sustainable ways.

He says supply chains should be economically sustainable for participants from farmers to intermediaries while protecting nature. He presents this as an aspiration and believes better development choices can support both livelihoods and the environment.

Asked what leaders should discuss more, Adams points to the changing influence of a growing organization. As Canva reaches more people, decisions and communication can affect employees, customers and the wider world. He says leaders need to think more carefully about that responsibility.

He recalls a brief conversation that later inspired someone to change a PhD focus toward financing biodiversity. The experience surprised him: words that feel casual to a leader can shape someone else’s decisions. Stebbings thanks him for the conversation.