Canva CTO: AI Agents Now Write Production Code Overnight
Canva's Chief Technology Officer revealed that the company is using AI agents to write production code overnight while its engineers are sleeping. This workflow allows human engineers to focus on higher-level tasks like design, architecture, and complex problem-solving during their working hours, significantly accelerating feature velocity.
- The role of senior engineers at Canva has fundamentally shifted from writing code to reviewing it. CTO Brendan Humphreys states their job is now "largely review," focusing on verifying AI-generated output and providing high-level direction to the AI agents. - This trend is not unique to Canva; Spotify's CEO Gustav Söderström noted that their most senior developers have not written any code since December, instead supervising an internal AI system. - Human oversight is critical due to what AWS CTO Werner Vogels calls “verification debt"—the time developers must spend reviewing AI output. A 2026 survey by Sonar found that 96% of developers don't fully trust AI-generated code, yet only 48% always verify it before committing. - While AI automates routine coding, demand is rising for engineers who can translate ambiguous product requirements into precise, machine-executable instructions. Humphreys stressed that effective use of AI agents requires "precision of articulation" and deep "mastery of the domain" to ensure correctness across Canva's 70 million lines of code. - The shift is creating a bifurcation in the engineering job market, with a Stanford study showing a 13% decline in employment for early-career engineers in AI-exposed roles, while senior roles remain stable or are growing. Gartner predicts that by 2027, this trend will require 80% of engineers to upskill. - Before wide-scale implementation, Canva's leadership took a measured approach to adoption by licensing multiple AI coding tools, such as GitHub Copilot and Cursor, and allowing its engineering teams to experiment and find what worked best for them. - Canva's existing engineering culture, which emphasizes collaboration, continuous learning, and clear communication, is well-suited to this new workflow. The interview process already assesses for skills beyond pure coding, such as how candidates communicate tradeoffs and contribute to team collaboration. - To manage quality, a robust peer-review process remains essential. A global survey initiated by Canva revealed that even as 92% of tech leaders adopt AI coding assistants, 93% ensure that all AI-generated code is reviewed by a human before being merged into production.
