5 Learnings Building Canva for AI Agents

2024-03-10 · Source: Engineering Leadership · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Robotics & Autonomous Systems · Depth: Intermediate, medium

Summary

Canva has transitioned its product development focus from traditional user interfaces to supporting AI agents, a shift highlighted by Anwar Haneef's talk at Engineering Leadership LIVE. This strategic move involves introducing AI features, an MCP Server for asset search, creation, editing, and publishing, and AI connectors for platforms like ChatGPT and Claude. The company observes three industry shifts: a behavior shift where agents call APIs instead of browsing UIs, a distribution shift positioning agents as a primary interaction surface, and a value shift where agents prioritize objective functionality over subjective aesthetics. Key learnings from this transition include treating the MCP setup as a living system, maintaining robust underlying APIs, acknowledging agents' non-deterministic and "picky" interaction patterns, prioritizing authentication as a core product feature, and emphasizing deep domain expertise to stand out to agents.

Key takeaway

For AI Architects and VPs of Engineering building products for the agentic era, recognize that AI agents are a new distribution channel demanding an API-first strategy. Prioritize Agent Experience (AX) over traditional UI, ensuring your underlying APIs are robust and debuggable. Invest in authentication as a core feature from the outset to avoid integration friction. Focus your product on deep domain expertise to become an indispensable tool for agents, rather than a generalist.

Key insights

AI agents represent a new distribution channel, demanding products shift from UI-centric design to API-first, prioritizing agent experience and specialized domain expertise.

Principles

• Agents call APIs, not UIs.
• Agents are a primary distribution surface.
• Objective agents prioritize function, reliability.

Method

Canva's approach involved building an MCP Server and AI connectors for platforms like ChatGPT/Claude, alongside revamping API design to support agent interaction and ensure debuggability.

In practice

• Design APIs for agent interaction.
• Prioritize Agent Experience (AX).
• Invest in robust authentication early.

Topics

• AI Agents
• Agent Experience
• API Design
• Product Strategy
• Authentication
• Domain Expertise

Best for: CTO, AI Product Manager, Entrepreneur, Director of AI/ML, AI Architect, VP of Engineering/Data

Related on AIssential

• I Kept Hearing About AI Agents Everywhere — So I Finally Sat Down and Learned What They Are — AI agents are goal-pursuing systems that reason, use tools, and operate in loops, fundamentally changing software interaction.
• Agentic Architecture Part 2: The Agentic & Shared Development Environment — Agentic platforms orchestrate specialized agents across diverse technologies to transform intent into action.
• Stop Getting Good at Protocols. Get Good at Agent Experience. — Agent Experience (AX) is a critical discipline for designing how AI agents interact with systems, transcending specific protocols.
• Karpathy: Stop Building for Humans — Future device interaction should prioritize agents using APIs over humans using individual apps.
• AIE Miami Keynote & Talks ft. OpenCode. Google Deepmind, OpenAI, and more! — AI agents are rapidly transforming software development, demanding new approaches to quality, efficiency, and architectural design.
• 5 Best Books for Building Agentic AI Systems in 2026 — Books offer depth and coherence for building agentic AI systems, surpassing fragmented online content.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Engineering Leadership.