# How Workday Builds Agent Prototypes in Under a Day with LangChain

## Executive summary

Workday, an HR and financials company, built an internal agent platform to streamline the development and deployment of AI agents for large enterprises. By partnering with LangChain, the team was able to drastically reduce the time required to move agent prototypes to staging—from 4 to 5 months to less than a day. The platform handles complex enterprise requirements like authentication and security by running agents as A2A and MCP servers, allowing internal teams to focus solely on defining the agent's functional graph (e.g., ReAct loop or deterministic workflow).

## Key takeaways

- Accelerated Agent Development Cycle: The platform reduced the time required to get an agent prototype to staging from 4 to 5 months to less than a day. This speed is critical for achieving goals of 50 to 100 agents in production and over a million active users by year-end.
- Built-in Enterprise Security: All agents on the platform run as A2A and MCP servers, with authentication and security requirements baked in. This allows internal teams to bypass complex infrastructure setup and focus only on the agent's logic and graph structure.
- Focus on Function, Not Infrastructure: The platform abstracts away infrastructure concerns, meaning developers only need to worry about the agent's function and how to build its graph (e.g., ReAct loop or deterministic workflow).

## Technical details

- Agent Deployment Architecture: The platform enables agents to run as both A2A (Application-to-Application) and MCP (Microservice Client Protocol) servers, ensuring that authentication and security are handled natively.
- Platform Components: The solution utilizes LangChain's Agent Server and LangSmith, providing an end-to-end package that includes observability and evaluation tools, which the team noted was essential for rapid development.
- Agent Graph Design: Developers are responsible for defining the agent's functional graph, which can take the form of a ReAct loop or a deterministic workflow.

## Practical implications

- For build engineers managing complex, regulated environments, adopting specialized platforms (like LangChain's) that handle core security protocols (A2A/MCP) can drastically reduce time-to-market for AI features.
- The shift in focus from building infrastructure boilerplate to defining functional workflows (ReAct, deterministic graphs) is a key pattern for modern enterprise AI development.
- Partnering with specialized AI providers can provide not just software, but a deep level of technical maturity and rapid delivery speed necessary for large-scale internal productization.

## Topics

AI Agents, Enterprise Architecture, LLM Deployment, Platform Engineering, LangChain, LangSmith Deployment, LangSmith, LangGraph, LangChain Academy

Source: https://www.youtube.com/watch?v=xM_BhmEyKAY
