BAA Building Agentic AI

The Books · A two-volume series

Two books, one running example, built for different readers.

Both volumes follow the same project, an internal IT helpdesk assistant, from a single prompt to a multi-tenant enterprise agent. Volume I builds the mental model. Volume II builds the system.

Volumes

2 · paired

Total chapters

33 · 14 + 19

Pre-book

Free · first 50 each

Format

PDF · download

Start with Volume I if

You can read code, but AI is new to you.

  • You have heard "tool calling, RAG, agent loop" and you are not 100% sure what each is.
  • You want to build a working mental model, not memorize current tools.
  • You want one chapter, one idea, no jargon, one running example.
  • You are preparing to make architecture decisions next quarter.

Read Understanding Agentic AI Systems →

Skip to Volume II if

You ship agents already and want the architecture.

  • You have a working agent in dev, and now you need cost ceilings, security review, and an on-call rotation.
  • You want gateways, routing, multi-tenant, FinOps, OWASP LLM Top 10.
  • You are preparing for a senior AI-engineering interview.
  • You want one running example carried from prompt to production platform.

Read Designing Enterprise Agentic AI Systems →

Volume I · For software people new to AI

Understanding Agentic AI Systems

From GenAI to Agentic AI, Explained Simply (A Beginner's Guide)

A plain-language introduction to agentic AI for software people who are new to AI. It teaches one idea per chapter, in clear language with a real example, building a single project (an internal IT helpdesk assistant) from a plain chatbot into a real, fenced, tested agent. By the end the reader has an accurate working mental model of the whole field, not a memorized snapshot of current tools.

What you will learn

  • How AI evolved through four eras, and how agentic AI differs from generative AI.
  • What a language model actually is, next-word prediction, and why it both works and hallucinates.
  • The three building blocks: prompting, RAG, and structured output and tool calling.
  • What makes something an agent: the think, act, observe loop.
  • Memory, MCP, agentic coding, and when multi-agent systems help versus just cost more.
  • How to make agents safe (guardrails, human-in-the-loop) and how to evaluate them.
  • A hands-on capstone: building a tiny working agent yourself.

Volume II · For engineers and architects

Designing Enterprise Agentic AI Systems

An Architect's Field Guide for Engineers

A field guide for engineers and architects designing agentic AI systems for real organizations, systems with cost ceilings, security reviews, and an on-call rotation. Across ten chapters it follows one running example, an IT helpdesk assistant, as it grows from a single prompt into a secured, evaluated, multi-tenant enterprise platform with a voice channel. Where Book 1 explains what an agent is, this book delivers the engineering judgment, architecture, and hard tradeoffs of running agents reliably at scale.

What you will learn

  • An accurate, load-bearing mental model of how LLMs behave in production.
  • Production prompting, structured outputs, function calling, and prompt registries.
  • Embeddings, vector search, and the full production RAG pipeline.
  • Agent design patterns, memory tiers, multi-agent tradeoffs, and bounded autonomy.
  • Reference architecture for an enterprise AI system: gateways, routing, services.
  • Cost, performance, and inference optimization, from token economics to caching.
  • Evaluation and quality: golden datasets, LLM-as-judge, agent evals, regression.
  • Reliability, security, and governance: OWASP LLM Top 10, audit, incident response.
  • Voice AI, LLMOps, multi-tenant platforms, FinOps, and Forward Deployed Engineering.
  • A full senior AI-engineering interview bootcamp.

Both books · One list

Get notified when each volume launches.

Pre-book either volume to be on the first-50 list for that book. Or just join the newsletter to hear when chapters land and when the books go on sale.