I Wrote Two Books Using AI — Here's What I Learned

Before I start, let me be clear about what "using AI to write books" actually means — because there's a version of that phrase that means "I asked ChatGPT to write a book and slapped my name on it," and a version that means "I used AI as a powerful tool to go from idea to published faster than any traditional process allows."

What I did was the second version. Both books — From McDonalds to Financial Freedom and The Autonomous Engineer — are authentically mine. The ideas, the stories, the framework, the voice. AI assisted with research, structuring, draft acceleration, and iteration. Here's exactly how.

The Workflow: From Concept to Published

Step 1: Brain dump first, AI second

I started both books with a raw dump of everything I knew about the subject — voice memos, notes, outlines, half-formed ideas. This is the irreplaceable human step. AI cannot generate your actual experience, your specific stories, your hard-won perspective. It can only help you organize and expand what you already have.

Skipping this step and starting with AI produces generic content that sounds like everything else. That's not what I was building.

Step 2: Use AI to build the skeleton

Once I had my raw material, I used AI to help structure it. I'd give it my brain dump and ask it to identify chapter groupings, logical sequences, gaps in the argument. This is similar to having a smart editor who reads your notes and says "here's how this could flow."

The AI was often wrong about what I wanted to emphasize. But it surfaced structural options I hadn't considered, and it was fast. Iterating on structure with AI took hours, not weeks.

Step 3: Write the first draft myself

I wrote the first drafts of every chapter. Some sections I wrote very quickly — the personal stories especially. Some sections needed research and more back-and-forth with AI to get right. But the voice in those first drafts was mine, because I was typing.

Step 4: Use AI to expand, tighten, and QA

After the first draft, AI became a revision partner. I'd paste in a section and ask it to: identify where the argument felt weak, suggest where I was being too vague, check whether the logic held, and flag anything that contradicted earlier sections.

This is where the 10x productivity multiplier came in. A good human editor can do this too — but they're expensive, slow, and not available at 2 AM when I'm in a writing sprint.

What AI Is Actually Good At (And What It Isn't)

The InkEngine Pipeline

For Book 2, I had already built a more mature AI writing pipeline I now call InkEngine. It handles: chapter brief generation, draft expansion, revision flagging, SEO metadata, and even book formatting for KDP submission. All command-line driven. All integrated with Claude's API.

The difference in speed between writing Book 1 and Book 2 was significant. Not because my writing got better (though it did), but because the tooling got better. Book 2 went from first outline to published faster than Book 1 went from first outline to first draft.

What the Future of AI-Assisted Writing Looks Like

The writers who will thrive are the ones who have something real to say and learn to use AI to say it more efficiently. The writers who will struggle are the ones who try to use AI to generate something they don't actually know or believe.

Readers aren't stupid. Generic, AI-generated content that exists only to fill pages or rank for keywords is already saturating the internet. What's scarce — and increasingly valuable — is a real human perspective, delivered clearly, with the friction of production removed by good tooling.

That's the combination I'm betting on. And so far, it's working.