How Workday Built an Enterprise-Scale AI Agent Platform
Workday — the HR, finance and IT platform serving more than half the Fortune 500 — used LangGraph for agent orchestration and LangSmith for tracing, observability and evals. A team of four built an internal developer platform that cut prototype time from 4–5 months to under a day, while satisfying some of the strictest audit, compliance and data-retention requirements in enterprise software.
Key numbers
- 4–5 months → under 1 day — time to build a prototype agent in staging, with authentication and security baked in
- 4 → ~30 people — size of the Agent Platform team
- End of 2026 — leadership's target for agents in production
- Over half the Fortune 500 — the scale of Workday's customer base
How the platform was built
- The platform is not fast for speed's sake — it is fast compliantly: authentication, security, compliance and infrastructure are handled by the platform, so teams focus on what the agent does and what the graph looks like. The first major product was the Agent System of Record, a governance layer controlling which APIs agents may access.
- Graph-model flexibility: compared with platforms that model agents as sequences of conversations or personas, LangGraph's graph model supports everything from simple ReAct loops to complex deterministic workflows, without the platform prejudging what an agent should look like.
- Protocols came free: LangChain's agent server implementations are built with A2A and MCP support, so when senior leadership asked about the strategy, the team could immediately answer that it was already in production. “The backend isn't the problem. LangChain took care of it for us.”
- An SDK unified internal capability: the team built “Workday LangChain”, a Python SDK wrapping LangChain primitives and routing them through internal services such as an LLM gateway and a RAG retrieval service. Internal services built over years in isolation became, through the platform, the first unified interface for building agents.
What Workday says
“The backend isn't the problem. LangChain took care of it for us.” — Alex Gatto, Senior Machine Learning Engineer, Workday
What it means for Hong Kong SMEs
When Hong Kong businesses choose an agent platform, they should not only ask how fast the demo is. Ask three things: permission governance (which systems can the agent use?), traceability (when something goes wrong, can you find why?), and compliance built in from the start (are audit requirements there from day one?). FinityAI helps clients design agent permissions and approval workflows starting from exactly these three questions.
Sources
This article is a FinityAI summary and commentary on a public article, and is not affiliated with Workday or LangChain.
Related services
SRAA services · AI Agent security readiness · AI Agent workflow planning