How Workday Built Its Enterprise AI Agent Platform with LangGraph

According to LangChain Blog:

Facing the demand to build and deploy production-ready AI agents by 2026, a small initial team of four at Workday set out to create an internal developer platform. Operating under strict enterprise security and data governance policies, they needed a solution that supported rapid prototyping and robust compliance. By leveraging LangGraph for agent orchestration and LangSmith for evaluation and observability, the team abstracted away underlying infrastructure.

They developed a Python SDK dubbed 'Workday LangChain' to wrap primitives and connect with internal systems like LLM gateways and RAG services. This setup allowed them to out-of-the-box support advanced protocols like A2A and MCP. As a result, prototype delivery dropped from five months to under a day, enabling the team to scale up significantly and unify Workday's fragmented internal services into a single cohesive agent interface.