Ivory Dog

From Conversation to Book

How to Explain Your Book Vision to an AI Writing Collaborator

Article 014 · About 11 minutes to read

In this article
  1. 1. What Is a Book Vision?
  2. 2. Begin With the Reader's Experience
  3. 3. Explain the Purpose Without Dictating Every Detail
  4. 4. Explain the Book's Intellectual or Creative Contribution
  5. 5. Describe the Voice You Want
  6. 6. Use Examples Instead of Relying Entirely on Adjectives
  7. 7. Communicate What the Book Must Not Do
  8. 8. Distinguish Facts, Memories, Ideas, and Speculation
  9. 9. Specify How You Want the Conversation to Proceed
  10. 10. Explain How Decisions Should Be Made
  11. 11. Use a Living Book Brief
  12. 12. Test the Vision Before Committing to a Complete Manuscript
  13. 13. A Complete Example of a Book-Vision Statement
  14. 14. Revisit the Vision During Revision
  15. Conclusion

One of the most important abilities an author can develop is explaining what kind of book they want to create.

This is not always easy.

A prospective author may recognize the book they want when they imagine it, yet struggle to describe that vision clearly.

They may know that the writing should be intelligent without being academic, personal without being sentimental, or entertaining without sacrificing accuracy.

Such distinctions are meaningful, but they require explanation.

When working with an AI writing collaborator, this task becomes especially important.

Claude can respond to written instructions, examples, questions, and feedback. But it cannot independently know what the author has in mind.

The objective is therefore not to produce one perfect prompt.

It is to establish a shared editorial understanding through conversation, examples, decisions, and revision.

1. What Is a Book Vision?

A book vision describes the intended work before all its details have been determined.

It should convey more than the subject.

A complete vision begins to establish the book's purpose, intended audience, structure, tone, intellectual contribution, and relationship with its readers.

Consider two descriptions.

Description A: I want to write a book about leadership.

Description B: I want to write a practical, evidence-informed book for owners of small manufacturing companies about making leadership decisions during financial uncertainty. The book should draw upon my own experience without pretending that every lesson applies universally.

The second description provides considerably more useful direction.

It identifies the audience, subject, approach, source material, and an important editorial limitation.

It does not determine the chapter structure or the final wording.

Those decisions can emerge through further discussion.

A useful vision provides direction without unnecessarily restricting development.

2. Begin With the Reader's Experience

Authors often describe books in terms of what they want to say.

That is necessary but incomplete.

The intended reader's experience should also be considered.

What should readers understand when they finish the book?

How should they feel?

What kind of attention should the book demand?

A novel may be designed to create tension, surprise, sympathy, or intellectual uncertainty.

A technical handbook may seek to make complicated procedures understandable and safe.

A memoir may invite readers to examine experiences unlike their own.

The author should communicate the intended experience to Claude.

Purdue University's prewriting guidance emphasizes the relationship between audience, purpose, and the choices writers make.[1]

For a book-length project, this relationship informs almost every editorial decision.

A chapter that works beautifully for specialists may confuse beginners.

A conversational style that suits a practical guide may be inappropriate for a scholarly monograph.

3. Explain the Purpose Without Dictating Every Detail

A vision should distinguish essential principles from negotiable choices.

For example:

This book must explain the subject accurately and acknowledge uncertainty. It should be accessible to readers without specialist training. I am open to different chapter structures.

The author has established three important requirements while leaving room for creative development.

By contrast, an instruction that specifies every chapter, example, heading, paragraph length, and transition before discovery begins may inhibit useful exploration.

Some detailed requirements are necessary.

A technical book may need to conform to particular terminology, units, standards, or documentation practices.

A novel may need to preserve established characters and events in a planned series.

The important point is to distinguish what must remain fixed from what may evolve.

Claude can then propose alternatives without repeatedly challenging settled decisions.

4. Explain the Book's Intellectual or Creative Contribution

Claude should understand what makes the proposed book distinctive.

For nonfiction, this may be original research, professional experience, a particular theoretical perspective, or access to unusual source material.

For memoir, it may be a personal viewpoint or an underdocumented experience.

For fiction, it may be a compelling narrative premise, unusual setting, distinctive characterization, or particular thematic concern.

Consider an aspiring historical novelist.

A weak instruction would be:

Write an exciting novel set in nineteenth-century England.

A stronger explanation would be:

I want a character-driven historical novel about a family whose economic security depends on a small textile business. The story should examine conflicting obligations between family loyalty, economic survival, and social ambition. Historical details should be researched, but the central characters are fictional.

The stronger version identifies the book's imaginative centre.

Claude can now ask more useful questions.

Which period of the century? What changes threaten the business? What does the protagonist want? What moral choices should the story explore?

The conversation begins to develop a particular novel rather than a generic historical narrative.

5. Describe the Voice You Want

Terms such as professional, literary, engaging, and intelligent can be interpreted in different ways.

An author should explain what those terms mean in the context of the proposed book.

A professional manual may require concise sentences, defined terminology, and a restrained tone.

A memoir may benefit from reflective prose, precise sensory details, and an understated sense of humour.

A thriller may require urgency, carefully controlled pacing, and economical description.

An effective voice description includes both positive and negative guidance.

For example:

I want clear, thoughtful prose with occasional dry humour. Avoid exaggerated emotional language, generic motivational statements, and elaborate metaphors that distract from the subject.

That instruction is more informative than simply requesting excellent writing.

Anthropic's prompting guidance recommends clear, specific instructions, sufficient context, and relevant examples.[2]

These practices are useful during book development because they make stylistic expectations easier to evaluate.

6. Use Examples Instead of Relying Entirely on Adjectives

An author may find it difficult to describe a preferred prose style.

Examples can help.

The author might supply a passage from their own previous writing and explain which characteristics should be preserved.

They might also describe the narrative qualities they admire in a particular kind of literature.

However, a request for a distinctive voice should not become an instruction to reproduce substantial copyrighted passages or closely imitate another author's protected expression.

The better approach is to identify general characteristics.

For instance:

I admire prose that is economical, observational, and emotionally restrained. I want careful detail and natural dialogue, but I do not want the text to mimic any particular writer.

This allows Claude to work toward the desired qualities while supporting a distinctive book.

The author should review sample passages before approving the general style for the entire manuscript.

Visual 1. Translating a book vision into editorial instructions
Vision elementVague directionMore useful instruction
AudienceWrite for everyoneWrite for beginners who have some practical experience but no formal training
PurposeMake it informativeExplain the causes of common equipment failures and how to investigate them safely
VoiceMake it literaryUse precise sensory detail and restrained reflection without melodrama
ScopeCover everythingFocus on decisions made during the first five years of business ownership
EvidenceMake it accurateIdentify important factual claims and require verifiable supporting sources
StructureOrganize it wellDevelop chapters from foundational principles toward practical applications
BoundariesDon't make mistakesDo not invent events, quotations, dates, technical specifications, or personal memories
CollaborationHelp me writeAsk one focused question at a time when important source material is missing
Original Ivory Dog instructional table. The examples demonstrate greater specificity without suggesting that any prompt can guarantee perfect AI output.

7. Communicate What the Book Must Not Do

Negative requirements can be as important as positive ones.

A memoirist may want to avoid revealing private family information.

A scientist may reject speculative claims presented as established conclusions.

A technical author may insist that hazardous procedures be flagged for specialist review.

A novelist may want to avoid certain plot conventions or stereotypical character portrayals.

These restrictions should be documented early.

They should also be specific enough to guide decisions.

For example:

Do not invent dialogue for real people. If a conversation cannot be reconstructed from notes or reliable recollection, summarize the event without presenting fabricated speech as fact.

Or:

Do not present my personal experience as proof that a practice is universally effective. Identify where research or technical standards are needed.

Such instructions help establish the author's editorial boundaries.

They do not eliminate the need to examine the resulting manuscript.

8. Distinguish Facts, Memories, Ideas, and Speculation

A long book project often involves several types of material.

It may contain firsthand memories, documentary evidence, original interpretations, creative proposals, and unresolved questions.

Claude should not treat all of them as equally established.

One practical method is to classify source material according to its status.

Verified: Supported by an identified, examined source.

Author testimony: Supplied by the author from memory or personal experience.

Creative proposal: A fictional or editorial possibility under consideration.

Interpretation: An explanation or inference that needs approval or support.

Unresolved: Information that remains unknown or uncertain.

These categories provide a foundation for reliable drafting.

A memoir chapter can use the author's recollections while acknowledging uncertainty.

A novel can incorporate invented scenes because their fictional status is understood.

A scholarly work should not adopt an unverified claim merely because it fits the argument.

Anthropic explicitly warns that Claude can generate incorrect information and convincing but unsupported statements.[3]

Clear source classification helps make these limitations manageable.

9. Specify How You Want the Conversation to Proceed

The author can also explain the preferred method of collaboration.

Some people want to discuss ideas freely before considering an outline.

Others prefer highly structured sessions addressing one chapter at a time.

An author with limited writing experience may prefer Claude to ask a focused question, examine the answer, and then ask an appropriate follow-up.

A specialist may prefer to supply lengthy technical notes and receive questions identifying missing explanations.

The method should suit the person and the book.

For example:

Begin by interviewing me about the subject. Do not draft chapters until we have agreed on the book's purpose and outline. Ask follow-up questions whenever an explanation is incomplete. Preserve my answers and distinguish them from your suggestions.

This establishes a procedural framework rather than simply describing the final output.

It is particularly valuable in long-form projects, where premature drafting can produce substantial material that later proves unsuitable.

10. Explain How Decisions Should Be Made

A book is the result of many decisions.

Which examples should be included?

Which chapter should come first?

How much background does the reader need?

Which interpretation of an event is most defensible?

Should a character behave differently in a particular scene?

Claude can help generate and compare alternatives.

But the author should determine which decisions require explicit approval.

A practical workflow might distinguish three classes.

Exploratory suggestions: Claude may propose possibilities freely.

Significant editorial changes: The author must approve alterations to structure, argument, narrative, or meaning.

Final manuscript decisions: The author reviews and approves the text before it is accepted as complete.

This arrangement allows collaboration without surrendering editorial authority.

11. Use a Living Book Brief

The initial vision should be recorded in a document that can evolve.

This document should capture the current approved concept.

It can contain the book's intended audience, central purpose, proposed genre, tone, boundaries, source materials, and major decisions.

When a conversation produces a significant change, the brief should be updated deliberately.

For example, an author may begin with a memoir intended for family members and later decide that the strongest material has wider historical significance.

That change may affect research requirements, privacy considerations, explanatory context, and publishing plans.

It should not happen invisibly through gradual changes in Claude's draft language.

A versioned brief makes the evolution of the project easier to understand.

12. Test the Vision Before Committing to a Complete Manuscript

Once a preliminary vision has been established, it should be tested.

For nonfiction, develop a sample chapter and determine whether the material provides sufficient evidence and explanation.

For fiction, create a synopsis, character profiles, and one or two representative scenes.

For memoir, produce a short narrative passage based on a documented conversation.

The sample should be reviewed for several qualities.

Does it serve the intended reader?

Does the language reflect the author's preferred voice?

Does the material have sufficient depth?

Are important factual claims supported?

Does the passage represent the kind of book the author actually wants?

The goal is not to polish a sample indefinitely.

It is to identify weaknesses before those weaknesses are reproduced throughout the manuscript.

If the sample reveals a problem with the underlying concept, the author should revise the brief.

If the problem is primarily stylistic, more precise guidance may be sufficient.

13. A Complete Example of a Book-Vision Statement

Consider a professional author who wants to create a book about operating a family-owned manufacturing business.

A useful initial statement might read:

I want to write a practical nonfiction book for people who own or manage small manufacturing companies. The central subject is making sound decisions when financial information is incomplete or future demand is uncertain.

The book should draw heavily upon my own experience, including mistakes, but it should not present personal anecdotes as universal rules. Where broader conclusions are necessary, I want research and reliable data.

The tone should be intelligent, direct, and occasionally humorous. Avoid motivational clichés, exaggerated claims, and language that makes ordinary business decisions sound heroic.

I would like the book organized around recurring management problems rather than a chronological history of my company.

Begin by interviewing me to identify the most useful experiences and lessons. Ask questions when facts or explanations are missing. Do not invent examples, financial figures, quotations, or events. We should agree on the outline before producing full chapters.

This statement is not a finished book plan.

It is the foundation of one.

Claude can now ask more specific questions about the intended audience, particular experiences, available records, and the book's possible structure.

The author can revise the vision as new possibilities emerge.

14. Revisit the Vision During Revision

Even a carefully planned book may change while it is being developed.

A nonfiction author may discover an important body of research that challenges the original argument.

A memoirist may recognize a theme that was not apparent during early interviews.

A novelist may find that a supporting character introduces a more compelling conflict than the initial plot.

Such discoveries should not automatically be resisted.

But they should be evaluated against the book's intended purpose.

Ask whether the change strengthens the work or merely introduces another interesting subject.

A living book brief can help make this judgment.

The final manuscript should be coherent even if its direction has evolved substantially from the original idea.

Conclusion

Explaining a book vision to Claude requires more than describing the subject.

The author must begin to clarify the intended reader, purpose, distinctive contribution, voice, structure, boundaries, and preferred method of collaboration.

Not every detail needs to be settled at the beginning.

Indeed, one of the principal advantages of conversation is the opportunity to discover a better approach through questioning and reflection.

The strongest instructions therefore combine clarity with flexibility.

They establish what matters while allowing worthwhile possibilities to emerge.

Claude can help organize ideas, ask questions, propose alternatives, and develop draft material.

The author evaluates those proposals and retains responsibility for the finished work.

A clear vision does not guarantee an excellent book.

But it makes an excellent book easier to recognize, develop, and defend.

References

  1. Purdue University Online Writing Lab. Prewriting (Invention) General Questions. https://owl.purdue.edu/owl/general_writing/the_writing_process/prewriting/questions.html
  2. Anthropic. Prompting Best Practices. Claude Platform Documentation. https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices
  3. Anthropic. Claude Is Providing Incorrect or Misleading Responses. What's Going On? March 16, 2026. https://support.claude.com/en/articles/8525154-claude-is-providing-incorrect-or-misleading-responses-what-s-going-on
  4. Purdue University Online Writing Lab. Why and How to Create a Useful Outline. https://owl.purdue.edu/owl/general_writing/the_writing_process/developing_an_outline/how_to_outline.html

Visual attribution

Visual 1, Translating a Book Vision Into Editorial Instructions: Original comparative table prepared for Ivory Dog. No third-party imagery used.