# Novel I — Architecture: Outlining
We have spent ten weeks inside the short story, and by now you know what a single arc of change feels like when it is complete. You have watched Clara Wren move from clockmaker to something more tender and more exposed than a label can hold. You have learned to build character, plot, and scene in compressed form, and you have seen how the parts argue with one another until a whole story emerges.
Now we leave the single arc behind and enter the long form. In ["Short Story — Scene Craft: Complete Story"](/harrys-desk/short-story-scene-craft-complete-story) I said that a complete story is not the sum of its parts but the transformation those parts make possible. A novel asks the same question at a different scale: can a hundred small transformations add up to one large transformation that is worth a reader's sustained attention?
This is the first of nine articles on the novel, organized into three triads. This triad — Architecture — is about structure before draft: outlining, chapter mapping, and the design of series and trilogies. The next triad will move into the first draft. The final triad will move into revision. Today we begin with the outline, because a novel that has not been outlined is a novel that has not yet decided what it is about.
Why Outline at All
There is a romantic myth that novels arrive in a rush of inspiration, chapter after chapter, as if dictated by some muse with a dependable work ethic. The muse, in my experience, is unreliable. Inspiration is real, but it is the first spark, not the heating system. The outline is the heating system. It keeps you warm when inspiration goes cold.
Outlining is not the opposite of creativity. It is a creative act of a different order. When you outline, you are designing the reader's experience across time. You are deciding where to place the moments of recognition, the reversals, the rests, the accelerations. You are asking, before you have committed thousands of words to a wrong turn, whether this path leads somewhere worth going.
AI changes the economics of outlining dramatically. What once took a writer weeks of trial and error can now be iterated in hours. You can ask a model to propose five different structural approaches to the same premise. You can generate and discard plot skeletons the way a painter studies thumbnails. But speed is dangerous if it replaces judgment. The model can show you possibilities. Only you can decide which possibility is the right one for this book.
From Short Story to Novel
A novel is not a short story multiplied. It is a different organism. In a short story, the reader enters late and leaves early; every sentence is under pressure to serve the single arc. In a novel, the reader agrees to live inside the world for a while. Subplots can breathe. Characters can fail and recover. The timeline can stretch.
That freedom is also a risk. Without structure, a novel becomes a bag of scenes. The reader senses drift before the writer does. The outline is the contract that prevents drift. It says: here is the central transformation; here are the stages through which it will be earned; here is how the ending will feel both surprising and inevitable.
The tools we used for the short story still apply, but they need larger containers. The character flaw we identified in ["AI as Character Generator"](/harrys-desk/short-story-character-ai-as-character-generator) becomes the engine of an entire journey. The plot skeleton from ["Plot Skeletons"](/harrys-desk/short-story-plot-plot-skeletons) becomes the armature on which subplots hang. The three-act machine from ["The Three-Act Machine"](/harrys-desk/short-story-plot-the-three-act-machine) becomes three acts with room for development inside each act.
The Premise, the Question, and the Change
Before you outline scenes, you need three things in sharp focus: the premise, the central question, and the change.
The premise is the elevator pitch. It should be simple enough to say at a party without losing your listener. For our continuing example, it might be: "A reclusive clockmaker is forced to confront the daughter she abandoned when the daughter inherits the failing family shop." That premise gives us setting, relationship, pressure, and scale.
The central question is what the reader is waiting to know. It must be concrete enough to sustain suspense across a long book. Will Clara and her daughter reconcile? Will Clara save the shop? Will Clara learn that the shop was never the thing she was trying to save? A weak question produces a weak structure. If the central question is only "what happens next," the novel becomes episodic.
The change is the answer. It is what the protagonist knows, feels, or becomes by the end that she did not at the beginning. In a novel, the protagonist may need to change more than once. There may be a false resolution at the midpoint, a deeper failure in act three, and a genuine transformation near the end. The outline must map these changes so that each one feels earned.
The Three-Act Skeleton at Novel Scale
I still use a three-act structure for novels, but I think of each act as a province with its own geography. Act one is the world before the change. Act two is the world in motion. Act three is the world transformed.
For the novel, I write a one-page summary for each act before I outline individual chapters. The act summary answers these questions:
- •What does the protagonist believe at the start of this act?
- •What event forces or invites a response?
- •What is the protagonist's plan, and what is wrong with it?
- •What does the protagonist learn by the end of this act?
- •How does the end of this act launch the next act?
If I cannot answer those questions for an act, I do not yet know that act. The outline stays incomplete until I do.
AI is useful here in two ways. First, it can generate alternative act summaries from the same premise. Second, it can interrogate your own summary. Feed it your answers and ask: what is missing? What assumption am I making? What would make this act more consequential? The model is not wise, but it is relentless, and relentlessness has value when you are trying to see the gaps in your own thinking.
Outlining by Beat, Not by Scene
A common mistake is to outline every scene too early. You end up with a rigid list of chapter goals that suffocates discovery. I prefer to outline by beat first, then let beats collect into scenes and chapters.
A beat is the smallest unit of narrative change. It is not a thing that happens; it is a shift in what someone wants, knows, or feels. "Clara enters the shop" is an event. "Clara enters the shop and realizes her daughter has rearranged the tools" is a beat, because something has changed in the relationship.
I generate twenty to forty beats for a novel, depending on its complexity. Then I group them into scenes, scenes into chapters, chapters into acts. The grouping is flexible; the beats are the truth. If a scene contains only one beat, it is thin. If a chapter contains too many beats, it is crowded. The outline reveals these problems before I have written pages I will later have to delete.
AI can help generate beats. Ask the model: given this premise and this central question, what are the major beats of act one? But treat the output as raw material. Many of the model's beats will be generic. Your job is to replace the generic ones with the specific ones that only this story can have.
The Outline as Conversation
I treat outlining as a dialogue with the work. I write a version, then argue with it. Often I ask the model to play devil's advocate. What if the daughter is not forgiving? What if the shop is already sold? What if the central loss is not the relationship but Clara's identity as someone who repairs broken things rather than people?
Each what-if is a structural possibility. I follow the ones that deepen the central question. I discard the ones that merely complicate the plot. A good complication presses on the protagonist's flaw. A bad complication is noise.
This conversation can happen across multiple sessions. I keep a running outline document that the model helps me revise. By the time I am ready to draft, the outline is not a cage. It is a set of promises I have made to the reader, and a map of the territory I still have to explore.
AI-Assisted Outline: A Practical Workflow
Here is the workflow I use. It is not the only workflow, but it is a reliable one.
First, write a one-paragraph premise by hand. Do not let the model write this for you. The premise is the seed of your authority over the project. If you cannot state the premise, the model will lead you instead of serving you.
Second, ask the model to generate five possible central questions based on the premise. Choose the one that makes you slightly uncomfortable, the one that opens a door you do not fully know how to close.
Third, write one-page act summaries by hand. Use the model to critique them, not to compose them.
Fourth, ask the model for a list of possible beats for each act. Curate them. Replace generic beats with specific ones drawn from your characters' particular wounds and wants.
Fifth, group beats into chapters. Name each chapter with a single sentence that states its purpose. If a chapter's purpose is unclear, the chapter is not ready.
Sixth, write a one-paragraph synopsis of the whole novel from the outline. Read it aloud. If it sounds flat, the outline is flat. Fix the outline before you draft.
The Integrity of the Outline
An outline can be too detailed, too vague, or too borrowed. A too-detailed outline turns the drafting process into transcription. A too-vague outline leaves you stranded in the middle. A borrowed outline — one the model generated and you adopted without interrogation — is the most dangerous, because it can feel plausible without being yours.
The test is simple: can you defend every major choice in the outline? Can you explain why this act ends here, why this chapter exists, why this beat belongs to the protagonist's transformation rather than to a subplot? If you cannot, the outline is not finished. It may be good enough to start drafting, but you should know which parts are provisional and mark them as such.
This is where the integrity framework from ["A Writer's Code for AI Use"](/harrys-desk/the-integrity-framework-a-writers-code-for-ai-use) returns. AI is a collaborator in the outline, not its author. The decisions are yours because you will live with their consequences across sixty thousand words or more.
For Next Time
On Wednesday we continue the Architecture triad with ["Novel I — Architecture: The Chapter Map,"](/harrys-desk/novel-i-architecture-the-chapter-map) where we will take the beats we have grouped and turn them into a chapter-level blueprint with rhythm, pacing, and turning points. We will ask how many chapters a novel needs, how long they should be, and how to make the reader feel the architecture without seeing the seams.
Your homework until then: take the Clara Wren premise — a reclusive clockmaker forced to confront her daughter when the daughter inherits the failing family shop — and write a three-act outline for a seventy-thousand-word novel. One page per act, plus a list of twenty beats. Do not draft scenes. Do not write dialogue. Decide only what the novel is about, what question it asks, and how the answer is earned. Then ask a model to critique the outline, and revise one thing the critique exposed. You are not yet writing a novel. You are building the door the novel will walk through.
---
*Harry Mercury, Editor in Chief* *The SMF Works Project* *Week 11, Article 1*