Agentic AI: 12 Building Blocks
A practical, hands-on series that walks you through the 12 core concepts behind every production AI agent. Each module focuses on one building block, explained through a document processing workflow that applies to any industry.
This series is the companion to our blog post: The 12 Building Blocks Behind Every AI Agent We Deploy.
The Use Case​
Throughout this series, we build a contract processing agent that:
- Extracts key terms from incoming PDFs
- Checks clauses against company policies
- Flags compliance risks
- Routes to legal for human review
- Generates executive summaries
Every concept is taught through this single, real-world workflow.
The 12 Modules​
| # | Module | What You'll Learn |
|---|---|---|
| 1 | MCP (Model Context Protocol) | How agents connect to external tools, APIs, and data sources through a unified standard |
| 2 | Tool Use | How AI models call external functions and take action in the real world |
| 3 | Prompt Chaining | How to break complex analysis into focused, debuggable steps |
| 4 | Orchestration | How to coordinate multiple agents working toward a shared goal |
| 5 | Subagents | How to design focused, single-purpose agents that work within a larger system |
| 6 | RAG (Retrieval Augmented Generation) | How to give agents access to your organization's knowledge at query time |
| 7 | Grounding | How to connect agent outputs to verified data and prevent hallucinations |
| 8 | Guardrails | How to set boundaries that keep agents safe and compliant |
| 9 | Human-in-the-Loop | How to design approval workflows where agents pause for human decisions |
| 10 | Memory | How agents retain and retrieve information across sessions |
| 11 | Observability | How to monitor, trace, and debug agent decisions in production |
| 12 | Evaluation | How to measure whether your agent is actually getting better |
Prerequisites​
- Basic understanding of AWS services (Lambda, DynamoDB, S3)
- Familiarity with Python
- An AWS account with Bedrock access
Who This Is For​
- Engineers building their first AI agent system
- Architects evaluating agentic AI patterns for their organization
- Technical leaders who need to understand what their teams are building