The Best Product Should Win: How I’m Evolving the Way I Write

My wife noticed first.

I had been playing with a tool I built to help me write. I call it Voice Muse. It was trained on over 400K words of my own work. Years of blog posts, chapters, speeches, and drafts of my first book. The idea was simple: make something that works with my voice rather than replace it. It takes my notes, sometimes a complete draft and cleans it up in my own voice. Often, at this point, I will read it and give the AI further direction as I would a copy editor. In the end, I do a line edit, and it becomes this blog.

I was experimenting with it on a blog post last year. I also asked Ed, a published author I had been working with, to take a pass at the same piece and gave him my notes and draft. Ed is good. Really good. I trusted him to improve my writing.

When my wife read both versions side by side, she said something I did not expect. The AI-assisted draft sounded more like me.

Ouch. What do I tell Ed?

I had spent months building a careful process. The whole point was to make sure the human product was the final product. I had planned to use my tools product as an input to Ed’s experienced pen. I wanted to point to a body of work someday and say I wrote this. No asterisks. No fine print.

Her comment messed that up. If a tool, trained on my own words, under my direction, through back-and-forth conversation, produced something that sounded more like me than what a skilled human collaborator produced, what exactly was I protecting?

Eventually, I decided I would lean into this tool for my blog. But my books would be human-produced.

My first book, Better Places, involved no AI. I did have a developmental editor. And while I am proud of what we wrote, I also felt that too much material got left on the shop floor. And in the end, I yielded to timelines rather than resolve a few structural issues that would have been easily navigated with my new tool.

The second one, Room to be Human, which I am writing now, is a hybrid approach that uses AI only to manage cognitive organisation, planning, and design. It will be a human product, and I am using an amazing ghostwriter, whom I have come to adore. The tool keeps the shop floor clean, though, and tracks any material I don’t use for later consideration. I am very sure that this book will be much better than my first one.

If I’m being honest, I’m starting to suspect this may be a more truthful reflection of my intentions and a more complete structural product. However, to humanise it, I may have to dilute it. And to meet my deadlines, I’ll have to compromise on flexibility. No offense to my editing team. They simply can not run as fast as I would if it were just me and my tool.

I recently asked Ed to likewise help me with a third book I am starting to develop.

The Email That Changed My Approach

A few weeks ago, Ed and I were talking about my next book. It is called Leadership Malpractice. I showed him my tool. I walked him through how I use it. I showed him the chapters I had already developed with it. His response surprised me. Here is a guy I was about to hire as a writing coach/ghostwriter, and this is what he wrote:

“I wonder if you need to be so concerned about whether the final product is ‘human’ or ‘AI’ or some combo. What I’ve seen in the AI drafts is very compelling writing. Writing that seems to reflect your unique story and viewpoints. Could you perhaps be similarly unapologetic in writing a book with AI help? And perhaps explain your choice in keeping with your view that AI unleashes creativity?”

He was not pitching me on using more AI. He was pointing out that my stated values did not align with my process.

I have been saying for months that AI, used well, amplifies human creativity. But if I am honest, I was building a process designed to hide that I was using it. That is not honest. That is not me.

Ed is right. And the fact that he was willing to say so, rather than quietly take the ghostwriting gig under the original terms, tells me he is the kind of person I want around this project.

So we changed the plan. Ed is not going to sand off AI fingerprints. He is going to help me do what good editors have always done. Strengthen structure. Flag stories that do not land. Polish sentences while keeping my voice. The tool stays. The process stays. The transparency is new.

The Calculator in the Desk Drawer

When I was in grade school, calculators were contraband. If a teacher caught you with one, it got taken away. The reasoning was that calculators would let you cheat. You would not really know math. You would just be pressing buttons.

That position held until it no longer did. Today, every student has a calculator. The SAT provides one. Math education did not collapse.

What changed is that teachers figured out what calculators were actually for. They do not replace mathematical thinking. They offload the grunt work so students can focus on the thinking. The confiscation era looks silly now. Not because those teachers were bad people. They were just reacting to a new tool with the wrong frame of mind.

Picture the same thing a few years later. An accountant in 1985 refused to use Excel. Real accountants work with ledger paper, he says. He runs every calculation by hand. He makes sure his work looks hand-done. No one can know he had help.

He does this not because the spreadsheet is less accurate. It is more accurate. He does it because his clients prefer ledger paper. They are suspicious that spreadsheet-assisted work is somehow cheating.

We would call that accountant a fool. The spreadsheet is a tool that amplifies expertise. It does not replace it. His knowledge of tax code and client situations is what creates value. Excel just lets him apply that knowledge at scale, with fewer errors, with more time for the judgment calls that matter.

That is what my tool does for writing.

My 400,000-word corpus is the ledger. Decades of thinking, arguing, reporting, reflecting. The AI is not generating content from a blank page. It is working within what I have already done. I seed it with human drafts. We go back and forth. I make every call about what stays and what goes. What is true? What matters.

The tool amplifies my voice and knowledge. It does not substitute for them.

If that is cheating, then calculators are cheating. Spreadsheets are cheating. Word processors are cheating. Printing presses are cheating. We would have to draw the line at scribes copying manuscripts by hand.

I could be wrong about this. Maybe there is something different about writing. Maybe the human hand on every sentence matters in a way that it does not matter in math. But in my case, for non-fiction at least, I do not think so. And I would rather find out the truth than hide it.

There Has to Be a Correct Way

The calculator story has a second lesson.

The answer was not “ban calculators forever.” It was also not “let students use them however they want.” The answer was to figure out the correct way to use them. When. For what. With what understanding of the math underneath.

AI in writing is at the confiscation stage. The instinct is to ban it. Flag it. Strip it out. Catch people using it. That instinct is going to age badly.

There has to be a correct way to write with AI. One that keeps what readers actually value. Honest voice. Earned expertise. Genuine thinking. One that uses tools that can amplify those things when used well.

Here is the distinction I think matters: It is not cheating if the tool is amplifying the authentic value the writer brings. It is cheating when the tool substitutes for a value the writer does not have.

That distinction is the work. Figuring out what separates amplification from substitution. What good disclosure looks like. What processes produce trustworthy results? What readers actually need to know. Those are harder and more interesting questions than “did AI touch this?”

Detection tools, as they exist today, are mostly binary. AI or human. Flagged or clean. That framing will keep failing because it is asking the wrong question.

And these scanners are by no means perfect. One recent trend in social media is to scan the US Constitution with GPTZero, one of the strongest scanners available. It will tell you with a 98% certainty that our founders used ChatGPT.

The useful question is not whether AI was involved anyway! The useful question is whether the writer brought real value. Whether the process was honest. Whether the result is what it claims to be.

That is a harder thing to build a tool for. But it is the thing worth building.

An Industry Figuring It Out in Real Time

I want to be fair here. The publishing industry is in a genuinely hard spot.

The volume of low-effort AI-generated content flooding the market is real. The pressure to maintain reader trust is real. Detection tools are unreliable. They produce false positives on human writing. They miss AI text. Writing good policy is genuinely difficult.

I do not envy anyone trying to figure this out. I have respect for the publishers, editors, and agents working through it in good faith.

I also think the field has not yet developed the vocabulary to distinguish meaningful cases. The person pasting a prompt into ChatGPT and the person using custom tooling trained on decades of their own work and research are not doing the same thing. But current conversations collapse them together.

That is not anyone’s fault. It is early. We will get there.

What I can do is be clear about my own practice going forward.

What I Am Evolving Toward

From this point forward, my writing practice will be openly disclosed.

My next book will be produced with transparency about the tool, the process, the corpus, and who did what. Readers will be able to see under the hood and judge the work on its merits.

That feels right. It is consistent with what I write about. Thoughtful human-AI collaboration is part of the future of knowledge work. Pretending otherwise in my own practice does not match my stated values.

Ed will be a collaborator on the next project. Not a traditional ghostwriter. Something closer to what good editors and co-thinkers have always done. Sharpen ideas. Spar with me. Strengthen structure. Preserve voice. Catch what I am too close to see.

The AI will be in the room too. A tool I built and direct.

My wife will keep telling me when something sounds like me and when it does not. She has been right so far.

I do not know exactly how this lands with readers. Some will appreciate the transparency. Some will reject the approach on principle. That is fine. The people who reject it on principle were never going to be persuaded by concealment. They would just have been reading under assumptions that were not true.

I would rather be honest and let readers self-select than be evasive and keep them under false pretences.

The best product should win. And the best product is one that the reader can evaluate with full information about how it was made.

That is the standard I want to hold myself to going forward. I suspect it is also the standard the industry will land on, once the tools and norms mature. I would rather help figure that out in the open than wait for someone else to settle it.