All labs
Lab 64
Capstone

Contract-Risk Agent — Architecture on an Agent

The finale. Take one enterprise AI agent — a strategic-sourcing & contract-risk agent — and evolve it live through every concept in the workshop, in the order you learned them. Start at a naive monolith and switch on microservices, async/event-driven, scale-out, fault tolerance, caching, security and observability — watching the cost, latency, throughput and CPU meters move, the architecture reshape, and an enterprise-readiness scorecard fill in. Flip the green/red lens to see the difference between the agent knobs every course stops at and the architecture layer that makes it production-grade. Each stage links to its deep-dive lab.

Walk one enterprise agent from a naive first draft to production, one decision at a time. Press Next and watch the single thing that changes at each step — and the one number that moves.
Step 1 / 8
Start

A naive agent — one process

The strategic-sourcing & contract-risk agent starts as one main.py: the LLM plus every tool — RAG, ERP, finance, legal, approval — and the database, all in a single process. It works in the demo. Then real traffic arrives.

Agent CPU under load96%one box, about to crash
👤 Human
reviews a vendor contract
🚪 API gateway
pass-through
🤖 Agent · LLM (ReAct)
monolith: LLM + all tools + DB in one process
🛠️ tools live inside the agent
RAG · ERP · finance · legal · approval
🗄️ Vector DB + ERP DB
single index
7 steps to production
What just happened