Case Study: Canva builds an agentic support experience with Langfuse

Canva handles 80% of support interactions with Langfuse

Canva, a visual communication platform with over 250 million monthly active users, sought to build a sophisticated AI support experience. The challenge was to develop and operate a multi-agent system capable of handling a massive volume of user queries across different tech stacks, specifically Java and Python, while ensuring high quality and enabling contributions from non-technical domain experts. They partnered with the vendor Langfuse to implement an end-to-end LLM operations platform.

Using Langfuse, Canva implemented a full suite of observability, prompt management, and evaluation tools. The solution provided deep tracing for debugging, versioned prompt management that empowered non-engineers to make updates, and a robust system for automated and human evaluation. This enabled a small team to build an AI support system that now handles 80% of user interactions. The results included a significantly improved AI output quality, faster iteration speed, and better user experiences at a lower cost.

Canva