JGSMWK005 - Building Enterprise-Ready AI Agents: From Prototypes to Production

Course Description

Getting an agent to run once is easy, but building one that engineering teams, security reviewers, and auditors can actually rely on is a very different craft. This Workshop helps you move beyond flashy demos to focus on the architecture, controls, and operating discipline required to make agentic workflows dependable in real enterprise settings. Using Claude Code as the development environment and the open-source project as a reference, we’ll take a practical look at what a production-grade agent looks like under the hood. We’ll explore explicit LangGraph state machines, small named nodes, and the essential plumbing that turns a prototype into a deployable system. We’ll cover the design principles behind enterprise-ready agents, including structured atomic workflows and human approval gates for high-risk actions. From there, we’ll move into the operational stack, examining the OWASP LLM Top 10 threat model and RAG architecture patterns. We’ll conclude with a hands-on blueprint exercise to design a state graph, tool permissions, and monitoring approach—providing a build-ready plan engineering teams can execute immediately.

To fully participate in the interactive components of this workshop, you will need to sign up for accounts at GitHub and at Claude Code (Pro or higher).

Benefits to the Learner

• Develop an Agent Fit Map to prioritize workflows, distinguishing between simple scripted automation and high-leverage agentic systems
• Architect end-to-end designs using the Orchestration-Operational-Ledger pattern to ensure repeatable, auditable, and engineering-ready AI workflows
• Construct a comprehensive Control and Risk Pack featuring permission gates and policy documents to prevent unauthorized autonomous actions
• Apply a reusable AI build kit to transform complex business processes into observable agentic systems with measurable quality
• Implement systematic monitoring and stop/escalate rules to dramatically reduce operational drift and ensure safe, human-in-the-loop oversight

Target Audience

• Platform engineers responsible for building and operating production-grade agentic systems within enterprise environments
• Technical product managers accountable for agent reliability, security, and deployment quality across the product lifecycle
• Founders and engineering leads tasked with moving working prototypes into governed, production-ready systems
• Security engineers and architects designing governance, observability, and threat mitigations for LLM-based workflows