How to Write a Book With AI as Your Editor (Nights and Weekends)
How to write a book while working full time: a realistic schedule, the exact AI editor prompts to paste, and what to never hand over to AI.
For Marketing, Managers and Product Management

How do you write a book with AI while working a full-time job?
Pick one book and write it in short, protected sessions (30 to 90 minutes, a few times a week). Write every word of the draft yourself. Use AI as an editor, not an author: to pressure-test your idea, tighten your outline, flag where chapters sag, and keep you on schedule. Finish a rough draft first, then revise in passes. The book stays yours because the sentences are yours. AI shortens the distance between a messy draft and a good one.
TL;DR
- Write the draft yourself and use AI for feedback, structure and accountability. A book in your own voice is the whole point, and AI-written prose is the fastest way to lose it.
- Do the schedule math first. At 500 words per session and four sessions a week, an 80,000-word draft takes about 40 weeks. Smaller sessions still finish; skipped weeks don't.
- Treat AI as several different editors: a developmental editor, a skeptical reader, a line editor and a project manager. Each needs its own prompt.
- Never trust AI for facts, quotes, citations or legal claims. It can state invented details with total confidence. Verify everything that is not your own experience.
- Check privacy and publishing rules before pasting your manuscript anywhere or sending it out. Your employer's material is off limits, and retailers and publishers have their own AI policies.
- Track progress somewhere visible. A simple log of words, chapters and next steps matters more than any writing app.
Why do working professionals stall on a book?
Most people who stall are not bad at writing. They write all day: decks, briefs, emails, strategy docs. They stall for three other reasons.
- The project has no shape. "Write a book" is not a task. Without a premise, a reader and a chapter list, every session starts with deciding what to do, and that is where a tired person quits.
- Feedback arrives too late or never. Friends say "I loved it." Professional editors are expensive and come at the end. For months you write without knowing whether the book works.
- The schedule assumes a life you don't have. Plans built on "I'll write for three hours every morning" collapse the first week a launch, a sick kid or a quarter-end hits.
AI helps with the first two. It can turn a vague idea into a structure in an evening and give you a reaction to a chapter in two minutes. It cannot fix the third. That part is a scheduling decision, covered below.
What kind of book are you writing?
The kind of book changes what you ask AI to do, and the length you're aiming for.
| Type | Typical draft length | What AI is most useful for |
|---|---|---|
| Nonfiction (business, how-to, ideas) | Roughly 50,000 to 80,000 words | Testing the premise, tightening the chapter structure, spotting repetition |
| Memoir | Roughly 60,000 to 90,000 words | Finding the through-line, flagging chapters that don't move the story |
| Novel | Roughly 70,000 to 100,000 words, varying by genre | Tracking continuity, pacing feedback, character consistency checks |
These are rough ranges; genres vary and publishers differ. If you're aiming at a traditional publisher, look up the norm for your specific genre. Nonfiction is typically sold on a proposal and sample chapters, while fiction is generally sold as a finished manuscript. Self-publishing has no gatekeeper but still rewards a book people can finish reading.
How do you build a realistic nights-and-weekends schedule?
Start with arithmetic, not ambition.
Total words ÷ words per session ÷ sessions per week = weeks to a rough draft.
For an 80,000-word book:
| Pace | Weekly output | Weeks to draft |
|---|---|---|
| 300 words, 3 sessions a week | 900 | about 89 |
| 500 words, 4 sessions a week | 2,000 | 40 |
| 1,000 words, 4 sessions a week | 4,000 | 20 |
Three rules make the plan survive real life:
- Set a floor, not a goal. "Open the file and write 200 words" is a floor. You'll often write more. On bad days the floor still counts, and the streak is what finishes books.
- Fix the slots in advance. Two weeknight evenings and one weekend morning, put on the calendar as appointments. Deciding each day whether to write is how books die.
- Plan a draft and a revision phase separately. Drafting takes the longest. Revision usually takes as long again, or longer. Budget for both so you don't hit "done" at month ten and discover you're halfway.
How do you use AI as a developmental editor?
A developmental editor looks at the big picture: is the idea clear, does the structure work, does the reader get what was promised. This is where AI is most valuable before you've written much, because fixing structure is cheap now and brutal later.
Give it your premise in your own words, then ask it to attack it.
Prompt: premise stress test
You are a developmental editor who has worked on many books in [genre: e.g. business nonfiction / memoir / literary thriller]. I'm going to describe a book I want to write. Do not flatter me.
My premise, in my own words:
[2 to 5 sentences]
My intended reader:
[who they are, what they struggle with or want]
Why I'm the person to write it:
[your experience, 2 to 3 sentences]
Tell me:
1. What is the single clearest promise this book makes to the reader?
2. Where is the premise vague, generic, or already done to death?
3. What three questions would a skeptical reader ask in the first chapter?
4. What would make this book distinctive rather than a copy of existing books in the category?
Ask me up to five clarifying questions before you answer if anything is unclear.
Answer its clarifying questions honestly. The answers are often your best material.
Prompt: outline builder (you supply the content)
Here is my premise and a list of everything I want to cover, in no particular order:
[paste your brain dump: stories, arguments, examples, chapters you already imagine]
Propose two different chapter structures for a book of about [word count] words. For each, give chapter titles, a one-sentence purpose per chapter, and what the reader should feel or understand by the end of it. Then tell me which structure you'd choose and why, and what's missing from my list.
Pick pieces from both structures, rearrange, and rewrite the outline in your words. The outline is a working document. Expect to revise it at the halfway mark.
How do you use AI during the drafting phase?
Mostly, you leave it alone. Drafting is where voice is made, and the temptation to ask for "a first pass" is strongest when you're tired at 9 p.m. Resist it for the prose itself. A few uses are safe and genuinely helpful, though:
- Getting unstuck. Paste a paragraph you can't finish and ask: "What question is this paragraph failing to answer?" Then write the answer yourself.
- Interviewing you. Ask AI to interview you about a chapter, one question at a time, then write from your own spoken answers. Many professionals find talking easier than typing from a blank page. Dictate your answers and use the transcript as raw material.
- Next-session prep. At the end of each session, write two lines on where you stopped and what comes next. Paste your notes at the start of the next one and ask it to turn them into a five-minute warm-up. This keeps a 45-minute session from being eaten by re-orientation.
- Continuity. For long projects, keep a running document of names, dates, terms and claims. Ask AI to check a new chapter against it.
Prompt: session warm-up
Here are my notes from my last writing session and my outline for the chapter I'm on:
[paste]
Give me: (1) a two-sentence recap of where I am, (2) the one thing this chapter must accomplish, and (3) three questions I should answer in my next 30 minutes of writing. Do not write any of the chapter for me.
How do you revise a draft with AI as your editor?
Once the rough draft exists, revision is where AI earns its keep. Work in passes, from big to small. Do not polish sentences in a chapter you might cut.
Pass 1: structure. Give AI one chapter at a time (long manuscripts exceed what a tool handles well in one go, and chapter-level feedback is easier to act on).
You are a developmental editor. Here is Chapter [N] of my [genre] book. The chapter's job, per my outline, is: [one sentence].
[paste chapter]
Do not rewrite it. Tell me:
1. Does the chapter do its job? Where does it succeed and where does it drift?
2. Which paragraphs could be cut with no loss?
3. Where do I assume knowledge the reader may not have?
4. Where is the energy highest, and where does it sag?
5. What is the one change that would improve this chapter most?
Pass 2: reader simulation. Ask it to read as your target reader and tell you where they'd stop.
Read this chapter as [specific reader, e.g. a mid-career manager who is skeptical of business books]. Mark the exact spots where you would skim, get annoyed, or put the book down, and explain why in one sentence each.
Pass 3: line editing. Now work on sentences, but with constraints so it doesn't flatten your voice.
Here's a passage from my book. Don't rewrite it. Instead, list: repeated words and phrases, sentences over 35 words, passive constructions that hide who is acting, and any clichés. Quote each problem and suggest a lighter-touch fix. Preserve my voice: [two or three adjectives, e.g. dry, direct, conversational].
Notice the pattern: you ask for a diagnosis, then you fix it. When you let AI rewrite whole passages, you get smooth, generic prose, and you lose track of which sentences are yours.
What should you never trust AI with?
- Facts, statistics, quotes and sources. AI can produce plausible but invented citations, studies and quotations. Anything you didn't witness or personally source needs verification against the original. If it matters, find the primary source yourself.
- Real people and legal risk. For memoir or any book that describes real people, AI cannot tell you what is defamatory or private. Ask a publishing lawyer if you have a real concern.
- Your voice. If feedback starts pulling every chapter toward the same polished, neutral tone, push back. A distinct voice is why someone reads your book rather than a summary of it.
- Judgment about whether the book is good. AI tends toward encouragement. Ask explicitly for the weaknesses, and ask a few human readers too. A trusted beta reader beats any model on "did this land emotionally."
- Confidential material. Do not paste internal documents, client information or anything covered by your employer's policies. Check your company's rules on outside writing and on side projects before you start, especially for business books drawing on your day job.
Also check how the tool you use handles your text. Many AI products have settings controlling whether your conversations may be used to improve their models. Read them before pasting a manuscript.
Finally, check the rules of wherever you plan to publish. Publishers, agents, contests and retail platforms have their own policies about AI-generated or AI-assisted work, and they differ. Using AI for feedback is generally a different matter than using it to generate text, but read the specific policy and disclose where required.
How do you stay on track for months?
Most unfinished books aren't failures of talent, they're failures of momentum. Three habits keep it going:
- Log every session. Date, words written, chapter, and one line on what's next. Seeing 14 sessions in a row is motivating, and a gap becomes visible before it becomes a quit.
- Do a weekly 10-minute review. Paste your log into AI and ask: "Based on this, am I on pace for my target date? What should I adjust?" Let it do the math and the nagging.
- Set milestones with dates. Outline done, first three chapters, half the draft, full draft, first revision, beta readers. Spread them out and tell one person who will ask about them.
If you miss two weeks, don't restart the plan. Return to the floor ("open the file, write 200 words") and rebuild from there.
What's a sensible order of operations?
- Weeks 1 to 2: premise, reader, and the stress-test prompt. Decide if the idea is worth a year.
- Weeks 2 to 4: brain dump, outline, two rounds of structure feedback.
- Months 2 to 9 or so: draft, in your own words, on the fixed schedule.
- Next 2 to 3 months: structural revision, reader simulation, cuts.
- Then: line edit, human beta readers, and a final read aloud.
- After that: decide on a path (agent or publisher, or self-publishing) and prepare what that path requires, such as a proposal, query or formatting.
Your pace may be slower or faster. The order matters more than the dates.
Could you build your own book-tracking tool?
Plenty of writers track a manuscript in a spreadsheet that slowly falls apart, or pay for a writing app with features they don't use. If you'd rather have something that fits how you work, Remy builds web apps from a plain-English description, so you could ask for a private manuscript dashboard: a page per chapter with status and word count, a session log, a progress chart against your target date, and a notes field for "what's next." Because it's a real app with its own sign-in and database, the data stays in something you control, and the code is standard TypeScript in a git repo you own. It won't write the book, and it doesn't need to.
Frequently Asked Questions
Can I write a whole book with AI?
You can generate text that is book-length, but you'd be publishing something that isn't really yours, and it will likely read as generic. Readers, agents and platforms increasingly care about this, and some have formal policies. The more rewarding and defensible approach is to write the draft yourself and use AI for feedback.
How long does it take to write a book working full time?
It depends on length and pace. At roughly 2,000 words a week, an 80,000-word draft takes about 40 weeks, plus revision. A sample plan runs about a year from idea to polished manuscript, though your pace may be slower or faster. Shorter books and more frequent sessions bring that down.
Is it okay to use AI to edit my book?
Using AI for feedback, structure and proofreading-style checks is widely common, but rules vary by publisher, agent, contest and retail platform. Read the policy of wherever you plan to submit or publish, and disclose AI use where it asks you to. Never paste confidential or employer-owned material into any tool.
Should I get a human editor too?
If you plan to publish, yes. AI is a useful first-line reader and available at 10 p.m. on a Tuesday, but it tends to be agreeable and can miss what a person with taste and market knowledge would catch. A human developmental editor or a couple of honest beta readers should see the revised draft.
What if I don't know what my book is about yet?
Start with a brain dump of stories, arguments and questions you keep returning to, then use the premise stress-test prompt. Ask AI to interview you one question at a time. Often the book emerges from what you naturally keep explaining to colleagues and friends.
How we source, write, and correct our articles. Editorial standards
Related articles

How to Make an AI Animated Explainer Video in Genspark

How Much Does an AI Short Film Cost? The $102,000 Breakdown

How to Start Freelancing on the Side with AI (Step by Step)
