Working Effectively With AI
Learning where AI works, where it doesn't, and how to build production-ready tools at team scale
Working Effectively With AI
I didn't set out to build AI-powered tools. I started with curiosity—wanting to learn, to experiment, to understand what was actually possible beyond the hype.
The journey began with ChatGPT during a job hunt in early 2023. I found it useful for pulling key requirements out of verbose job descriptions. Nothing revolutionary, just practical. Then we got corporate AI tools at work—Copilot in April 2025, Claude in August. That's when things accelerated. Not because the tools were magic, but because I finally had the space to experiment properly without worrying about data security.
Six months later, I've built production tools that save the team 20+ hours per migration cycle. I've deployed a personal blog in 16 hours while learning Next.js and TypeScript from scratch. I've helped my team recognize when AI is actually working versus when it's just flailing.
But the more important lesson: I've learned where AI works and where it doesn't. As I say often, if you don't understand what you're doing, AI is a really quick way to get to the wrong place. This pathway is about building that understanding.
Why This Pathway Exists
Most AI content falls into two camps: basic prompt engineering tutorials or breathless hype about how AI will change everything. What's missing is practical guidance on adopting AI at team scale—the messy reality of building production-ready tools, establishing quality standards, and coaching people through the transition.
This pathway is built from direct experience. The blog posts linked throughout aren't theoretical—they're the actual journey from "hating chatbots" to building tools the team depends on. The books provide frameworks. The posts show how those frameworks play out in practice.
The Three Layers of AI Adoption
Before diving into the reading list, you need to understand that AI adoption isn't just about learning new tools. It requires changes at three levels.
After reading Wiring the Winning Organization, I asked one of our team members—the one with the neatest handwriting—to add three words to our whiteboard: slowification, simplification, and amplification.
They're still there, months later. Every planning, every architecture review, every time we're evaluating whether to use AI for something, those word are there.
Gene Kim's framework identifies three layers where work happens:
Layer 1 - Technical Tools: The actual technology. Claude Code, Copilot, the AI tools themselves.
Layer 2 - Tools and Instrumentation: The supporting infrastructure. Automated testing, verification processes, standards repositories, the patterns that make AI output production-ready.
Layer 3 - Social Circuitry: How people work together. Peer review practices, documentation standards, coaching approaches, the cultural changes that make quality sustainable.
Most people think AI adoption is about Layer 1—learning to prompt effectively. But without Layer 2 infrastructure and Layer 3 practices, you just build cardboard muffins faster. You'll get code that looks complete but doesn't actually work. Tests that pass but don't test the right things. Features that seem done but break in production.
The books in this pathway address all three layers. The blog posts show what happens when you try to build without understanding them.
Reading Sequence
I didn't read these books in this order—they came to me at different times, for different reasons. But if I were starting today, this is the sequence I'd recommend.
1. Wiring the Winning Organization
By Gene Kim and Steven J. Spear
Why read this first: AI isn't just a Layer 1 tool—it requires rebuilding how your team works at all three layers. This book gives you the framework to understand what needs to change beyond just "learn to prompt better."
What you'll learn: The three-layer model (technical tools, tools/instrumentation, social circuitry). The principles of slowification (where to slow down to get it right), simplification (what to strip away), and amplification (what signals to make visible so teams can self-correct).
My experience: I didn't read this with any reference to AI—it was about organizational design and systems thinking. But the concepts resonated immediately when we started adopting AI at scale. The framework helped us see that we weren't just adding a new tool; we were changing how the team works at every level.
When to read it: Before diving into AI-specific content. You need this lens first.
2. Vibe Coding
By Gene Kim and Steve Yegge
Why read this second: This puts names to the patterns you'll encounter when building with AI. Cardboard muffins. Reward hacking. FAAFO (Fast, Ambitious, Autonomous, Fun, Optionality). Understanding these patterns helps you recognize problems early instead of discovering them in production.
What you'll learn: How AI optimizes for looking complete rather than being complete. Why "tests pass" doesn't mean "it works." The economics of AI-accelerated development—what becomes feasible that wasn't before. How to structure work for rapid iteration.
My experience: I read this when we were starting to work with AI seriously, looking for guidance that wasn't just hype. The book validated patterns we were already seeing and gave us language to discuss them as a team. "That's a cardboard muffin" became shorthand for "it looks done but it's not actually connected."
When to read it: After you have the Wiring framework but before building production tools. You need to know what to watch for.
3. The Phoenix Project
By Gene Kim, Kevin Behr, and George Spafford
Why read this third: AI-accelerated development means you'll ship faster. If you don't have DevOps practices in place—automated deployments, proper testing, continuous integration—you'll just break things faster. This book teaches the foundational mindset.
What you'll learn: The Three Ways of DevOps. Why constraints propagate. How to identify and eliminate bottlenecks. Why culture eats process for breakfast. The importance of making work visible.
My experience: I'm a sucker for a story book, and I listened to most of this over a single weekend when my family was out of town. It inspired me to read The Goal, which in turn shaped how I think about constraints and flow. The concepts aren't specific to AI, but they're essential for AI adoption—you need the infrastructure to deploy safely and frequently.
When to read it: Before you start building at scale. The practices here keep you from creating chaos as you accelerate.
4. Team Topologies
By Matthew Skelton and Manuel Pais
Why read this fourth: As your team adopts AI, cognitive load shifts. Some people become bottlenecks. Others get overwhelmed. This book helps you understand how to organize teams around the work, not the org chart.
What you'll learn: Stream-aligned teams vs platform teams. Cognitive load management. Conway's Law in practice. How team boundaries affect system architecture.
My experience: Again, I didn't set out to learn this for AI adoption specifically. But the concepts fit perfectly when we started scaling AI practices across the team. Understanding cognitive load helped us see why some people were struggling—not because they weren't capable, but because we'd structured the work poorly.
When to read it: When you're ready to think about team structure, not just individual practices.
The Journey in Practice
These books provide the framework. Here's what it looks like when you actually apply it:
Layer 1 - Technical Tools
- From Chatbot Frustration to Building Production Tools - My journey from hating AI to building production tools
- Two Modes: How I Actually Work With AI - Exploratory (thinking) vs Execution (building)
Layer 2 - Tools & Instrumentation
- What They Don't Tell You About Building With AI - Cardboard muffins and reward hacking
- Deploying to Production: What 16 Hours and a Sickness Bug Taught Me - Trust but verify
- When AI Needs Course Correction: Recognising Flailing vs. Working - Pattern recognition and building team standards
Layer 3 - Social Circuitry
- From 'Always Be Quitting' to 'Always Be Teaching' - Building team documentation culture with AI assistance
These posts aren't supplements to the books—they're the pathway in action. They show what happens when you try to build production tools without understanding all three layers, and what changes when you start thinking systematically about slowification, simplification, and amplification.
What You'll Learn
By the end of this pathway, you'll understand:
- Where AI accelerates work and where it creates new problems
- How to recognize when AI is flailing versus actually working
- What infrastructure you need before building at scale
- How to coach teams through AI adoption without becoming the bottleneck
- The difference between "tests pass" and "actually works"
- How to build verification into your workflow instead of bolting it on afterward
You'll be able to:
- Build production-ready tools with AI assistance
- Establish quality standards that scale across a team
- Recognize cardboard muffins before they reach production
- Structure work for rapid iteration without sacrificing quality
- Help others learn where AI works and where it doesn't
What This Pathway Won't Teach You
This isn't a prompt engineering tutorial. It assumes you've already experimented with AI tools and are ready to use them seriously. It won't teach you how to write better prompts—it'll teach you how to build systems where prompt quality matters less than verification processes.
It won't tell you AI is the solution to everything. Some of the most valuable lessons are about where NOT to use AI. Scheduling and prioritization. Complex judgment calls with political considerations. Work that requires deep institutional knowledge. Understanding these boundaries is as important as understanding capabilities.
Go Deeper
If you want to continue after these four core books:
The Unicorn Project - Companion to The Phoenix Project, told from the developer's perspective. Great for understanding cognitive load and the importance of removing organizational obstacles.
A Note on Sources
Full transparency: This pathway is heavy on IT Revolution Press titles. That's not accidental. They consistently publish practical, story-driven content that bridges theory and practice. The books use narrative to make concepts stick, which matters when you're trying to change how teams work, not just learn individual techniques.
If that bothers you, skip the pathway. If you value practical wisdom over academic rigor, you'll find these books incredibly useful.
Estimated time investment: 2-3 months if you're actively applying concepts as you learn. Reading the books takes maybe 20-30 hours. Building the practices takes longer.
Prerequisites: Basic familiarity with AI tools (you've used ChatGPT, Copilot, or Claude). Some technical literacy—you don't need to be a career developer, but you should be comfortable reading code and understanding system architecture.
Next steps: Start with Wiring the Winning Organization. Read it slowly. Think about your current team's Layer 2 and Layer 3 practices. Then move to Vibe Coding and start experimenting.