AI Search
When Melanie Perkins, Cameron Adams, and Cliff Obrecht launched Canva 13 years ago, the goal was to make design accessible to everyone. Back then, it was a territory reserved for specialists. Adobe’s suite was the industry standard—powerful, but complex and expensive enough to keep most people out. Creating a polished presentation or a simple marketing flyer meant either hiring a designer or wrestling with tools built for professionals.
Fast forward to today, and the three co-founders are making their most ambitious move yet, redefining the entire creation process—and they’re using agentic AI to make it happen. On Thursday, Canva introduced Canva AI 2.0, its next-generation creative operating system capable not only of generating visuals but also of retaining knowledge of your organization and connecting with third-party apps to produce a range of project materials on your behalf.
“The community has been wishing for this for a long time,” says Obrecht, the company’s chief operating officer, in a press conference last week. “They’ve been asking for more AI, more deep workflows, helping them move from just creating a single piece of content—which we do really well currently—to really helping them achieve their whole job.”
It’s an opportunity that Canva isn’t alone in chasing. Adobe has been pushing Creative Cloud deeper into AI-powered content generation with Express and Firefly, while enterprise platforms like Microsoft 365 Copilot, Slack, and Glean have made AI content creation a core part of their pitch. But those platforms are largely solving for one part of the problem—generating content or automating tasks. With the exception of Adobe, most are largely not in the creativity business. Canva is making a bigger argument: that the entire process of creation is broken, fragmented across too many tools and too many workflows.
But what’s the difference between design and creation? The former is the output, such as a social media post, a presentation, or a flyer. Think of it as the last mile of a creative task. The latter, on the other hand, is the entire process, from the initial idea and research to drafting, formatting, collaborating, scheduling, and publishing. It’s the workflow, and this is what Canva wants to transform.
Thanks for reading The AI Economy! Subscribe for free to receive new posts and support my work.
Our mission at Canva has always been to empower the world to design. Design has never been about making something look pretty. [It’s] about giving form to an idea.
—Melanie Perkins, Canva chief executive
Canva first introduced Canva AI in May 2025, billed as an “all-in-one creative partner for generating designs, images, and text from a single prompt.” But it was largely a set of features integrated into a design tool. Canva AI 2.0 builds on that foundation with four new architecture layers—new ways to design with Canva—and five intelligent workflows—do more within one app.
Here’s what’s new:
The era of the traditional user interface appears to be drawing to a close. In an AI-filled world, we’re going to be engaging with apps using natural language, not clicking buttons. Canva is embracing that trend, and now you can use text or voice to describe what you want created, and it’ll be brought to life. It’s not just in the idea-creation phase, but throughout the entire design process.
“You can start a goal or a task like writing a business proposal, and Canva AI will figure out how to bring that to life,” Perkins explains. “And everything, very importantly, creates a fully editable design so you can jump into the editor at any time to edit an image, change a headline, or collaborate with your team.”
This feature empowers Canva AI to continuously act on and refine designs across a series of steps, rather than responding to a single prompt. Think of it as a take on vibe designing —Canva AI autonomously handles creative tasks on your behalf, allowing you to focus on everything else.
Running a marketing campaign? Canva AI can automatically create an Instagram post, Instagram story, YouTube thumbnail, and other creative assets, deciding on images, text, fonts, colors, and positioning without requiring constant instruction.
Canva AI is now intelligent enough to learn from your designs and adapt itself to your style and preferences, thanks to the new memory library. Think of it like the system prompt you’d give an AI agent, except Canva builds and updates it automatically based on how you design, so you never have to write it yourself.
The memory library is built on top of Canva Docs, meaning everything stored in it is fully editable, so if the AI gets something wrong, you can correct it like any other document.
The first book that Canva supports is called “About Me.” It captures your personal design style and preferences, building a profile of you based on your existing designs. And it will continuously update as you design, collecting more information about you. Canva says more books are coming, though it hasn’t specified which ones or when—Perkins claims the company will be launching more “over the years to come.”
The potential here is significant. Imagine separate books for individual clients, project-specific context, or company departments. Each one has its own brand guidelines, goals, and creative history. For freelancers managing multiple clients, or enterprise teams juggling dozens of campaigns, that kind of persistent, organized memory could fundamentally change how creative work gets handed off and scaled.
There’s a practical consideration worth noting too: For anyone using Canva across personal and professional projects, the books structure could be what prevents proprietary client work from bleeding into unrelated designs. It’s certainly something that’s easy to overlook until it becomes a problem.
Powered by the Canva foundation design model that the company launched in October, the layered object intelligence feature turns static, flat images into fully layered, editable designs from the start. Every element—images, text, colors, positioning, and fonts—is individually accessible and adjustable, just as if you’d built the design from scratch yourself.
It also means Canva AI can edit individual elements without needing to “regenerate” the entire design. For example, ask it to rewrite a headline and only that headline changes—everything else stays where it is. It’s a feature that Adobe has had in Firefly and Photoshop since 2025 with layered image editing. For anyone who has lost hours to regenerating an entire composition just to tweak one detail, this is probably going to be a blessing.
Powered by Anthropic’s 3.7 Sonnet model, Canva Code is a tool to create landing pages, countdown timers, games, dynamic forms, and other interactive applications using HTML and JavaScript. Using natural language prompts, it generates code that can be embedded into web pages or used within Canva’s platform.
Today, the company is unveiling support for HTML Importing, a feature that makes any HTML or code file brought into Canva fully editable in the drag-and-drop editor. Want to edit a button color in the HTML? You no longer have to regenerate the entire file—individual elements can be edited independently, similar to how Layered Object Intelligence works with design assets.
Context is key, and Canva wants to bring more of it to its platform to help you create more impactful work. Canva AI 2.0 comes with direct integrations into apps like Slack, Gmail, Google Drive and Calendar, Notion, HubSpot, Atlassian, Linear, Microsoft 365, OneDrive, and Zoom.
“Canva AI connectors enable you to go out to different sources and incorporate those right into your Canva designs,” Adams, the company’s chief product officer, says. That means you can ask it to prep a document for an upcoming meeting, and Canva AI can extract the details straight from your Google Calendar—no copy-pasting required.
Third-party connectors follow a familiar pattern in enterprise software—reduce context switching, keep data where work happens. But for Canva, it’s also a strategic move: the more of your workflow that lives inside Canva, the stronger its case for being the system at the center of how creative work gets done.
Not all the context Canva AI needs lives inside your organization. Sometimes it’s out on the internet, and that’s what the web research feature is for. Say you ask Canva AI to generate a business plan for your wellness and nutrition startup—web research will go out and find relevant resources, and pull them into a fully formatted Canva Doc. That research becomes the foundation for whatever you need to build next, be it a pitch deck, marketing campaign, or brand brief.
In a way, this scheduling feature puts Canva AI on autopilot. You specify the task and the deadline and let AI take the wheel. You can instruct it to pull the latest updates from Gmail and Slack, compile them into a branded newsletter, and have it ready every Monday morning, all without you needing to do any work.
Brand consistency has long been one of the hardest things for growing organizations to maintain. As AI-generated content becomes more prevalent in enterprise workflows, staying on-brand by default rather than by correction is becoming a competitive necessity. Brand Intelligence addresses that directly.
“Canva AI 2.0…goes from manual formatting to iterative agentic editing,” Perkins says. “It goes from having dispersed brand guidelines to brand intelligence…from having scattered data all over the web to connectors that pull it all together…” Borrowing a phrase from one of Canva’s engineering leads, she describes the result as “an orchestra where everything works together seamlessly.”
You can certainly use each feature individually, but Canva’s argument is that the whole is greater than the sum of its parts — and that this has been the architectural bet it’s been building toward for more than six months.
Canva AI 2.0 is rolling out today as a research preview, but it is limited to the first one million people who discover it. While the company showed attendees of its annual Canva Create event how to get early access, you won’t have to suffer from FOMO. Here’s a hint on how you can jump the line: type in “activate superpowers.”
Along with this new platform, Canva says it has been steadily investing in developing its own multimodal foundation models designed for design. And while the company’s first design model took two years to build, these new models were much faster, taking as little as a month to complete.
That said, here are Canva’s three new proprietary models, which the company says are optimized for output quality, speed, editability, and real-world workflows:
Canva Proteus: Built for style transfer and is 2x faster and 23x cheaper than comparable frontier alternatives.
Canva Lucid Origin: Built for image generation and is 5x faster and 30x cheaper.
Canva I2V: This is an image-to-video generation model and is 7x faster and 17x cheaper.
Canva also announced a deeper integration with Anthropic’s Claude. According to the company, both the Canva Design Engine and its Visual Suite are now accessible directly within Claude — and the connection runs both ways. Artifacts generated in Claude, ChatGPT, or other AI platforms can be brought into Canva, where they become instantly editable and publishable as fully functional websites.
All of these announcements are designed to further cement Canva’s Creative Operating System—its vision for a unified platform where AI, design, and productivity converge. The first version debuted last October with the launch of the company’s design model, its @AskCanva assistant, and an expanded Visual Suite. Canva quickly discovered that having its own design-aware AI model in place allowed it to iterate and build a more sophisticated agentic layer faster than it had anticipated.
That said, Adams acknowledges that while the design model made design generation better, “interacting with the designs themselves through AI was still a bit hard.”
Unlike its predecessor, this next-generation Canva AI is made to be a “full end-to-end creative partner” present at every stage of the creative process. It’s no longer a background utility but an active contributor. It doesn’t matter whether you start your own design or ask Canva AI to generate one; it knows not only what you’re currently creating but what you’ve produced in the past, drawing on all of that context as it works alongside you.
Adams also says that Canva AI 2.0 is not about creating designs, but about workflows. Design tools are ubiquitous thanks to generative AI and vibe designing, but the real value comes from thinking about what you do next with the synthetic work—how it gets assembled, branded, and deployed into something that actually moves a business forward. That’s the problem Canva bets it can solve.
Canva isn’t the only one chasing this opportunity. As generative AI continues to flood the market with synthetic content, platforms like Adobe Express are staking out the same territory, positioning themselves as the essential layer where raw AI output gets refined, branded, and made deployable.
People are using tools like ChatGPT, Claude, Gemini, and Copilot to generate information, but that’s only one part of the process. Next, you have to turn it into something visual that appeals to the audience—the data needs a home. Will it be a presentation, a branded PDF, a video, or a webpage? This presents a unique opportunity for Canva, Adobe, and Figma, all of which are racing to be the platform where this transformation happens.
But Canva believes its scale sets it apart. With 265 million monthly users, 31 million paid subscribers, $4 billion in annualized revenue, and a ranking among the top three most-used AI products in the world, it has both the reach and the resources to own that finishing layer—and the mission to bring it to everyone.
Share
No posts
