New Agentic Tools from Canva Transform Creative Design

Ever wanted to simply describe a vision and watch it come to life, ready to publish, in seconds? At its flagship Canva Create event, Canva unveiled Canva AI 2.0, an architectural overhaul that makes that wish a reality. 

The update allows users to describe a goal, an idea, or a rough outcome and watch the platform orchestrate the entire design and execution process through a single conversation.

This shift into conversational design marks Canva’s transition from a suite of creative tools into a fully agentic system. 

By leveraging a proprietary orchestration layer, Canva AI 2.0 no longer just generates static images; it interprets intent, selects the necessary tools and iterates on structured, multi-page assets via natural language dialogue.

The rise of the design agent

The update allows users to use agentic AI as a creative partner that understands context. 

A single prompt, such as “plan and design a multi-channel launch for our new eco-friendly shoe line”, triggers the AI to generate everything from social media assets to internal strategy documents, all while maintaining brand consistency.

To support this, Canva introduced several core intelligence features, including AI models that generate flattened images where every element is its own editable layer. 

Users can ask the models to swap the background but keep the product lighting the same, allowing for precise, non-destructive editing.

“We believed that design was something that should be accessible to everyone, not just a privileged few,” says Melanie Perkins, Co-Founder and CEO of Canva, writing in a Canva blog post, discussing how her business developed. 

Plus, Canva AI 2.0 features persistent memory, learning a team’s specific brand guidelines, preferred styles, and past choices to ensure every generated output feels proprietary rather than generic.

Extending generative capabilities to data, the tool builds structured spreadsheets, like project timelines, populated with real-world research and data points based on a simple description.

High-performance from low-latency infrastructure

Canva also revealed its Vertical AI stack. By self-hosting its training and optimising inference infrastructure, the company claims its proprietary models are significantly more efficient than current frontier alternatives.

These efficiencies allow Canva to scale generative AI to its 250 million monthly active users without the prohibitive costs typically associated with large-scale model deployment.

Deepening the AI ecosystem, even offline

Canva is positioning itself as the “visual layer” for other AI leaders. 

It has expanded its partnership with Anthropic to bring its Design Engine directly into the Claude interface. Users can now import “artifacts” from ChatGPT and Claude directly into Canva, turning raw AI text or code into fully functional, collaborative websites and presentations with a single click.

Canva has also developed connectors for Slack, Gmail and Google Drive. These workflows allow the AI to perform background tasks, such as scanning a Google Calendar to pre-generate briefing documents or summarising Slack threads into visual reports.

Additionally, the launch of Canva Offline ensures that these high-compute workflows remain accessible on mobile and desktop regardless of connectivity, with automatic cloud-syncing once back online.

Cameron Adams, Co-founder and Chief Product Officer at Canva, says: “We’re building for the small business owner in Chicago, the student in Manila, the community organiser in rural Brazil, people with real creative work to do who need AI woven into a workflow they already know, not a new technical discipline to learn from scratch.”

The Canva AI 2.0 Research Preview launches today. The rollout begins with the first one million users to engage with the update on the Canva homepage, with a progressive global release scheduled for the following weeks.