An open manuscript covered in handwritten editorial notes, with a fountain pen and writing materials arranged on a wooden desk.

Writing Begins Long Before the First Sentence

1st Draft / Revision| AI & Writing Tools

October 12, 2026•25 min read

What 15 years of unfinished manuscripts, a famous novelist's outlines, an 1874 typewriter, and ten months building my own AI writing assistant taught me about who really writes a book.

By Christopher di Armani

Fifteen Years Asking the Same Two Questions

For approximately 15 years, I could not complete a single first draft, despite starting dozens of them.

I had ideas. Tons of them. I wanted to write. Just like I wanted to breathe. That desire didn't go away when the next book idea stalled out. So I'd start over. I'd pick the next idea, make another attempt, always with the insane belief that this time would be different.

But sooner or later, the same old questions would whisper in my ear.

“I don't know what to write.”

“I don't know what to write next.”

That was my kryptonite. I didn't have a clear, detailed outline that showed me where the book needed to go and what each chapter needed to accomplish.

Every time I sat down to write, I had to figure out what the book was about, instead of writing it.

I didn't have a plan. I didn't have a writing schedule either. I wrote when I felt inspired. And while that inspiration could get me started, I could never count on it to get me back in my writing chair tomorrow, next Tuesday, or after a rough week.

For years, those two problems reinforced each other. When I sat down, I didn't know what came next. Not with any certainty. And because I didn't have a writing routine, it was easy to avoid dealing with uncertainty until I felt inspired again.

Today, it's no mystery why all those first drafts stalled out. I know what my recurring complaints mean, and I know how I approached writing.

Then one day I sat down and made the decision to finally try something that I knew in my bones would never work.

I wrote an outline. It wasn't a good outline, but it was a place to start. A place to finally answer those questions that plagued me for over 15 years.

I finally completed a first draft. And it was because I finally tried the thing I knew in my heart would suck the creativity out of writing, that thing that would never, ever work. One crappy outline.

That was the start of my development of the process I now call The Di Armani Method for Non-Fiction Writers.

I learned to make all of the important decisions before trying to turn every idea into polished paragraphs. I embraced the notion that the purpose of a first draft is to get the complete idea out of my head and onto the page.

I built detailed outlines that gave every chapter a specific job. I established Writing Appointments and learned to keep those appointments to write “whether I feel like it or not.”

The change transformed me from a writer who couldn't complete a single first draft in over 15 years to a writer who later wrote and published seven books in 90 days.

That's not a promise that following my system will produce the same result for you. It's evidence that the system worked for me, even under immense time pressures, so long as I worked the system.

I didn't discover a way to make writing every book easy. I developed a system that made writing every book simple. That repeatable method made the work ahead clearer, more manageable, and far more likely to be completed.

I spent too many years trying to write without knowing what to write next or, at times, what to write at all.

This history, I hope, makes it clear why I'm so passionate about writing from a detailed outline, writing on a schedule, writing whether I feel like it or not. And now, writing and the responsible use of artificial intelligence.

I'm not interested in returning to my old way of writing simply because someone believes that “a real writer” must struggle through every part of the writing process with no assistance at all, be it from a human or a machine.

The Outline Tells Me What to Write. The Scheduled Appointment Tells Me When to Write.

Some writers resist outlining because they think it limits creativity. That was me, for far too long.

Because of my experience, I see why a bare list of chapter titles might leave that impression.

If your entire outline says things like, “Chapter Four: Building Confidence,” it's of no help to you.

You still must decide what that chapter is responsible for teaching your reader. The label alone doesn't tell you which argument to make, what evidence supports that argument, or what your reader should understand when they reach the chapter's end.

A useful outline goes far deeper.

Before I outline a book, I need to understand its central message, its right-fit reader, the problem it solves for that reader, the promise I can responsibly make, and its intentionally limited scope.

Then each chapter needs a defined purpose within that reader journey.

What does this chapter teach?
What evidence or experience supports it?
What example belongs here?
What parts belong somewhere else?
What should the reader know or be ready to do next?

Once I make those decisions, I don't need to rediscover the book every time I sit down to write. I just open the outline, find the next assignment, and start.

That's one half of the change that helped me write and publish over 50 books.

The second half is a Scheduled Writing Appointment.

I put a specific time on my calendar. I decide what I'm going to work on, and I show up whether inspiration joins me or not.

And, as H. Jackson Brown Jr. so wisely said:

Don’t waste time waiting for inspiration. Begin, and inspiration will find you.

Of course, real life interferes at times. Family members get ill and need support. An emergency happens that rightfully forces you to pull off Publication Highway for a time.

So a functional writing system must include ways to recover from a missed writing appointment. That's also non-negotiable.

The important point is that inspiration no longer gets to decide whether I make time to write or not. Writing time is built into my calendar.

But an outline without execution is an elaborate plan that never becomes a book. A calendar appointment without a clear writing assignment can put me right back in front of the same old question: “What do I write next?”

Those two practices address the problems I know firsthand. One tells the writer where the next section belongs. The other protects the time required to write it.

Neither one removes the need to think, exercise your human judgment, or revise the plan when the need arises.

This distinction eventually became part of the Roadmap for Writing Success I teach through The Di Armani Method for Non-Fiction Writers.

Now consider what happens when someone else helps prepare the first draft of your book.

What James Patterson Actually Does

James Patterson is a useful example of an author who doesn't approach writing a book as a solitary undertaking.

In a 2014 interview with Noah Charney for The Daily Beast, Patterson explained that, on some collaborative projects, he prepares outlines of roughly 60 to 80 pages. Each outline addresses most of the chapters before his co-writer ever sees it. Patterson revises the outline, invites the collaborator to contribute to the plan, and then Patterson reviews manuscript pages every few weeks.[1]

Look at that detail.

He doesn't hand over a vague idea, disappear for a year, and put his name on whatever comes back.

Patterson's process involves discussion, evaluation, correction, and sometimes substantial changes of direction. Patterson has described polishing the completed collaborator's draft and, depending on the book, writing several more drafts himself.

His co-authors deserve their proper place in this account. Many receive clear public credit. They're not anonymous typists. They're true collaborators. They contribute ideas, expertise, judgment, and original creative work.

Was this always his system? No. But I'm willing to bet his emphasis on writing a detailed outline before starting a first draft is the foundation of his success.

Patterson does a remarkable amount of thinking before the collaborator produces the finished first draft. He establishes the direction of the book and a working structure that both writers can assess. He stays involved as the manuscript develops. And the first draft doesn't end the process.

For a nonfiction writer, the detailed outline looks very different. I'm not plotting a thriller or arranging fictional cliffhangers.

I'm deciding how to explain an idea, serve a specific reader, support a particular claim, and move that reader through a clear sequence. But the structural principle transfers perfectly.

If I know precisely what a chapter must accomplish, I can give a more useful assignment to a human assistant, an editor, or an AI writing tool. That outline also gives me a strong basis for judging whether the resulting chapters do their job.

Patterson's success doesn't prove that all delegated writing is good or that a machine is the equivalent of a human collaborator.

It does challenge the idea that authorship means you must personally type every sentence of every first draft.

It's helpful to remember that this isn't the first time a change in writing technology prompted questions about what counts as real writing.

Imagine the Objection in 1874

In 1874, E. Remington & Sons put the Sholes & Glidden Type Writer on the market. According to the Smithsonian's National Museum of American History, it was the first mechanical writing machine to achieve commercial success. It wasn't the first typewriter ever invented, but it established the typewriter as a practical writing tool.[2]

Now, imagine someone watching a writer use one of those newfangled machines and declaring, “You're not a real writer. Real writers use pen and paper.”

That imagined declaration exposes a poor way of judging authorship. The tool used to put words on a page may change, but the person who uses the machine still develops the argument, chooses the words, and decides what the finished manuscript ultimately says.

A typewriter didn't make a poem less of a poem or a letter less personal. A computer keyboard didn't take away the writer's ideas. The machine changed the physical act of recording language. That's all.

But generative AI introduces a complication that the typewriter or personal computer never did.

A typewriter records the words its operator chooses. An AI tool can propose words, sentences, arguments, transitions, and entire drafts. It can supply expressive material, not merely reproduce the author's keystrokes.

We must acknowledge that difference, not cover our ears and pretend it doesn't exist.

Here's the point I want to make: adopting a more capable writing tool doesn't make you less of a writer. What matters is the ideas you contribute, what decisions you delegate, what creative control you exercise, and how honestly you represent the finished work.

Which leads to another part of this story. My current AI writing assistant didn't become useful because I typed one clever instruction into an empty chat window. Far from it.

Ten Months Building an Assistant That Works the Way I Write

I've spent approximately ten months building and refining what I now call The Di Armani Method Operating System. It's a customized ChatGPT Project, developed through countless rounds of writing prompt instructions, testing AI's responses, finding the problems, correcting them, and establishing ever more precise standards for the way I write.

I didn't create or train the underlying AI model. I developed the working environment that tells the model how to assist me.

Most of that work never appears in the final manuscript. The reader sees a paragraph. They don't see the months I spent figuring out what information the system needed before that paragraph became possible.

For example, I developed a detailed personal voice, style, and tone guide. It describes the way I approach a reader, the language I favor, the sentence patterns I use, and the habits I want removed when a draft starts to sound like generic AI prose.

I didn't expect the system to guess how I write simply because I ask it to “write in my voice.” I gave it standards and examples, and I continue to evaluate what it produces.

I developed a personal Story Bank containing real experiences I've used in my writing and teaching over the years. Those accounts provide details that a language model cannot legitimately invent on my behalf. If a personal story belongs in a chapter, the system needs an actual story, not an impressive-sounding fabrication.

I also established a theological governance system. I'm a Reformed Baptist who adheres to the London Baptist Confession of Faith of 1689, which is always subordinate to Scripture.

Those convictions govern how I approach biblical interpretation and doctrinal claims.

An AI writing assistant needs to know those boundaries if it's going to help with my Christian nonfiction writing. It must not quietly replace my beliefs with a generic religious message or invent biblical authority for an idea I happen to like.

Then there's the writing method itself: my book-foundation decisions, outlines, Roadmap for Writing Success, editorial checks, and rules for what the system can and cannot do.

I revise this set of instructions (the 25 files that make up the Operating System) when I discover they produced something misleading, shallow, overly polished, or inconsistent with my actual position.

All of that took concentrated effort over time. Real work. Months of time.

I made decisions about my language, my ideas, my experience, my theology, and the standards for acceptable writing. I developed materials that didn't exist in that form before I created them. I tested the output against what I know I mean, and I changed the system when it missed the mark.

Improving the operating system, like writing itself, is an ongoing creative process.

Today, the system produces writing that comes quite close to how I write. It's not perfect, but it's not meant to be.

I already built my right-fit reader, writing method, voice, and editorial standards into the operating system. I don't need to explain all of that every time I want to write something.

What I do need to supply is the substance of the particular article, chapter, or book we're working on. That includes my ideas, arguments, sources, firsthand experiences, and the decisions about what the writing needs to accomplish.

But that information doesn't always arrive in one neat package.

Sometimes I begin with an idea and we work through it together. I question the AI's suggestions. It challenges my assumptions. I discover something important I overlooked. I add an experience, correct an interpretation, or change the outline. Some of those decisions happen before the first draft. Others emerge as we develop and revise the manuscript.

The AI helps organize that material and turn it into a working first draft.

Then I take over.

I examine what it produces. I challenge weak reasoning, verify factual claims, remove anything it invented, restore details it missed, and rewrite language that doesn't sound like me or doesn't say what I intended.

That's what I need from my AI writing assistant: a useful first draft that I can examine, challenge, and rewrite, not a machine-produced manuscript I feel obligated to accept.

I can't give you a controlled numerical voice-match score or claim that my process will save you the same amount of time. I can only show you what I've built and how I use it.

And I challenge the assumption that none of this preparation counts as work by the author.

It plainly counts as intellectual, editorial, and creative-direction work. That doesn't automatically make every sentence the machine generates legally mine in every jurisdiction.

It means the visible first draft only tells part of the story of how that draft came to exist.

A Human Co-Writer and an AI Assistant Aren't the Same Thing

This is where my argument needs precision.

I began this project with a much stronger claim. I suggested that using AI to write a first draft was, in substance, no different from hiring a human writer to prepare it.

The comparison has value, but it isn't the right one.

A human co-writer can bring independent knowledge, life experience, editorial judgment, and original creativity into the work. A collaborator may deserve author credit and have rights or responsibilities established by contract. Patterson's collaborators, for example, can do far more than follow instructions.

An AI writing assistant can also take an assignment, generate a preliminary draft, respond to corrections, and help reorganize material. Those functions overlap with parts of a human collaborator's work.

But software has no firsthand memories to contribute as human testimony. It can't stand behind a claim as a human author, and has no human co-author's contractual standing.

If I gloss over those differences, I weaken the case for responsible AI assistance.

The defensible argument is much simpler.

Writers have long used other people and increasingly capable tools to help bring their books into existence. That assistance doesn't, by itself, settle who authored the finished work.

We must examine what the person named author genuinely contributed.

In my case, we don't need to invent an example to see what that contribution looks like.

We already created one.

This Article Is the Demonstration

Look at the process behind the piece you're reading right now.

I began with an idea about James Patterson. I wanted to explore his approach to outlines and co-writing, then compare that approach with a writer who uses AI to help create a first draft.

I believed the analogy could challenge a common objection to AI-assisted writing.

But I didn't ask the AI to write a glowing defense of my position. I asked it to conduct a Red Team Review first: to test my assumptions and tell me what I was missing.

The response challenged several parts of my original premise. It pointed out that Patterson uses credited human co-authors, not simply interchangeable ghostwriters.

It raised questions about what constitutes original authorship, how substantially revised AI-generated text is classified by publishing platforms, and whether a polished voice match actually represents the writer's thinking.

Those weren't ideas I personally supplied in the opening prompt. The AI contributed analysis that made the argument more demanding and, ultimately, stronger.

I could have ignored those objections. Instead, I used them to improve the plan.

Then I asked it for a detailed outline.

The outline was massive, but it left out important pieces of my own history, like my 15 years of unfinished first drafts.

I noticed the omission and said so. Without that correction, the article would have begun with Patterson instead of the experience that explains why I teach outlining with such conviction.

Next, I clarified what actually stopped me during those years. I supplied my precise internal language: “I don't know what to write” and “I don't know what to write next.”

I added the second problem, my habit of waiting for inspiration rather than protecting a scheduled Writing Appointment.

The assistant reorganized the article to distinguish structure from scheduled execution.

I also introduced the 1874 typewriter comparison. The assistant checked the date and improved the argument by distinguishing a mechanical typewriter from an AI system that can propose original-sounding prose. The historical analogy came from me. The qualification that makes the analogy more accurate emerged through our exchange.

Later, I realized the outline still missed the ten months I spent developing my customized AI environment. I explained the voice guide, theological governance, Story Bank, and other foundational work.

That changed the article again, because it exposed the human preparation that takes place long before the first sentence appears.

We even worked over a single word. The assistant proposed a five-sentence thesis. I wanted the second sentence to refer to “that first sentence,” not merely “that sentence.” I instructed the system to add the word “first,” and that correction became part of the working thesis.

Finally, I recognized something the assistant hadn't identified, and it was central to the illustration: my entire exchange with my AI operating system was the writing process in microcosm.

I supplied ideas, tested them, spotted omissions, corrected language, and decided what the article needed to become.

The AI OS questioned, researched, organized, proposed, and revised.

That's collaboration. It's not a story I made up to defend my position. It's the record of how this article actually developed.

But notice what it doesn't prove.

It doesn't prove that every word of the original first draft came directly from my keyboard. It didn't. That was the point of having AI assist me with the first draft.

What the record now shows is that I've taken that first draft and substantially rewritten it in my own words.

I changed the language, strengthened the arguments, added firsthand experiences, and made the manuscript my own.

I spent approximately five and a half hours developing this article, from our initial discussions and outlining through my extensive rewrite.

I spent about two hours working back and forth with my AI assistant on the outline. The rest I spent revising the manuscript.

What it doesn't establish is precisely how much time this process saved compared to writing the article entirely without AI.

What the record also shows is how much human thinking and decision-making took place before the first draft ever existed. And that creative work didn't stop when the AI delivered its first draft.

Without my contributions, this would be a completely different article. And much of its most important source material wouldn't be here at all.

The First Draft Isn't the Whole Test of Authorship

Here's the working principle that emerged from this collaboration. I asked for one word to be changed, and I agreed that the resulting statement captured the direction I wanted for this article:

An author's work doesn't begin when the first sentence appears on the page.
It begins with the thinking, research, experience, decisions, and preparation that make that first sentence possible.
Sometimes that preparation includes a detailed outline. Sometimes it includes developing an entire writing system that carries the author's voice, knowledge, convictions, and editorial standards.
AI can now assist with producing the first draft, just as human collaborators have assisted professional writers for generations.
But the author must still contribute the original creative substance of the book, make the decisions that shape its expression, and take responsibility for the final finished manuscript.

Those are important distinctions.

The work of developing my AI system is genuine labor. Some of that labor involves original writing and intellectual organization. But developing a useful assistant isn't identical to creating original expression in every paragraph of every future book.

For a specific manuscript, I still need to ask where its ideas, evidence, structure, and expressive choices came from.

Did I originate the claim?

Did I supply the firsthand experience?

Did I select the evidence and decide what it means?

Did I choose the right-fit reader and each chapter's job?

When the draft was wrong, did I correct its reasoning or simply smooth out the sentence?

Which passages reflect my own original expression, and which remain generated text that I accepted?

That last question can be uncomfortable. But it needs to be asked anyway.

I place enormous value on line-by-line rewriting.

When I revise, I can reconstruct arguments, reject words I wouldn't use, add genuine experience, change emphasis, remove unsupported claims, and make the manuscript express what I actually believe.

But changing every adjective isn't the same as supplying the book's intellectual substance.

A person could ask AI to invent the subject, argument, examples, and chapter sequence, then replace enough words to make the result look different.

That process isn't the same as bringing an original idea and a developed body of knowledge to the writing desk.

On the other hand, a writer can make significant original contributions through research, structure, personally written sections, selection, arrangement, and substantive creative revisions even when some preliminary text comes from AI.

There's no honest shortcut here. We must examine the actual work and what each contributor supplied.

And we must divorce our ethical assessment of authorship from questions about legal copyright ownership and publishing-platform rules.

The Objections We Shouldn't Dodge

Start with the strongest one: “A typewriter doesn't write sentences. AI does.”

Absolutely. That's why the typewriter comparison is a limited analogy, not a declaration that both technologies work the same way. It challenges the idea that a new tool automatically disqualifies the writer. It doesn't resolve the separate problem of identifying original human expression in AI-generated prose.

Another objection is that Patterson's collaborators deserve credit. Of course. Their creative contributions must not disappear just because his name appears prominently on the cover. My argument about AI doesn't require diminishing the human writers who collaborate with him.

Then there's the objection aimed directly at my own system: “Ten months spent prompting AI isn't the same as authoring a particular book.”

Building an author-specific system is substantial intellectual and editorial work. Authorship of any particular finished manuscript also depends on my creative contribution to that manuscript.

One can't automatically stand in for the other. That's precisely why I keep original source material, detailed outlines, and authorial revision central to the process.

What about copyright?

The U.S. Copyright Office explained in its 2025 report on AI copyrightability that copyright protection can apply where a human determines sufficient expressive elements, including certain human-authored material, creative arrangements, or modifications. Mere prompts aren't enough by themselves. The Office also states that using AI in the creative process doesn't automatically bar copyright protection for the human-authored material.[3]

Canada's 2025 government report on a consultation about generative AI and copyright documents the policy questions around authorship, ownership, and liability. It's a report on what participants said, not a sweeping new rule that resolves every AI-authorship case.[4]

There's also a practical publishing distinction that writers can't ignore.

Amazon KDP's published content guidelines distinguish between AI-generated and AI-assisted content.

If an AI tool creates the actual text, KDP classifies that text as AI-generated even after substantial human editing.

If the human creates the text and uses AI to improve or check it, KDP treats that differently. KDP requires disclosure of AI-generated content but not AI-assisted content, under those guidelines.[5]

Those classifications matter regardless of whether I consider my creative contributions meaningful. Editing a draft doesn't give me permission to ignore a distributor's definitions.

A separate objection concerns authenticity.

My voice guide may help the AI assistant approximate my writing style, but resemblance alone doesn't make a claim true.

The system can produce a sentence that sounds like me and still get my experience, my beliefs, or my research wrong.

That's one reason I always read and challenge its output instead of treating AI's fluency as evidence.

And what about speed? I've already told you about ten months of preparation. That's an investment, not a shortcut.

A generated first draft can appear quickly, but it still needs verification and revision.

To claim that my particular system saves a specific amount of total time, I'd need comparable measurements from complete writing and revision cycles. I don't have them and I'm not claiming otherwise.

Finally, writers must handle confidential material, third-party rights, and reader expectations responsibly. A personal Story Bank doesn't give an AI permission to invent a testimony. A private client interview doesn't become public source material simply because software can process it. And whatever a platform requires us to disclose, we still have an obligation to describe our contribution honestly.

These objections are worth answering because they protect something I care about deeply: the credibility of the author and the trust of the reader.

Back to the Writing Desk

For 15 years, my questions were painfully simple.

“I don't know what to write.”

“I don't know what to write next.”

A detailed book outline gave me an answer. A scheduled Writing Appointment gave me a time to act on it. Those changes helped me move away from repeated failed first drafts toward a series of completed and published books.

Now I use another tool. I spent months preparing it to work with my ideas, my experience, my established writing process, and my editorial standards. I still make the manuscript decisions. I still review the words. And when the draft doesn't say what I mean, I change it.

Patterson's collaborative process reminds us that substantial preparation and creative direction can coexist with delegated drafting.

The typewriter reminds us that changing the tool doesn't remove the writer.

AI adds new capabilities and new responsibilities, and I'm happy to address both.

You don't have to use AI to be a serious writer. You can write every line yourself, work with a human collaborator, or use carefully bounded software assistance.

Whatever route you choose, you still need something worth saying and a clear plan for saying it.

Here's what I'll ask you to do today.

Open the outline for one unfinished chapter. If it doesn't exist, create it. Decide what that chapter must help your reader understand, which original knowledge or experience you can contribute, what evidence belongs there, and what the next writing assignment is.

Then put a Writing Appointment on your calendar and, most importantly, show up for that appointment.

If an AI assistant can help you carry out a well-defined assignment without replacing your judgment or inventing your message, decide whether that assistance serves the book. If it doesn't, write the passage yourself.

Either way, you still have work to do.

The work begins long before the first sentence. The tools may change. But you still have to know what you're saying, decide how to say it, and own what you publish.