Most writing workflows track motion. A card slides across a board while the audience, the argument, and the evidence stay exactly as vague as they were on the first day. A workflow earns its place only when each step ends in a decision. What follows is a six-stage sequence for long-form work, with the test that closes each stage.
- Qualify an idea by naming the reader, their question, and your answer.
- Fit the brief on one page and treat non-goals as the load-bearing field.
- Stop research when every listed question has an answer or a named gap.
- Read the outline sentences aloud in order to test whether they argue the case.
- Split revision into structure, evidence, clarity, and voice passes.
Stage one: qualify the idea
An idea is not a project until you can say who needs it and what changes for them. Write three lines: the reader, the question they arrive with, and the answer you can give. If the third line is guesswork, the idea is not ready to schedule.
Keep a shelf on purpose. Most ideas are not wrong, only early, waiting on evidence you do not have. The test for leaving this stage is narrow: you can state the reader and the promise in one unhedged sentence each.
Stage two: settle the brief
The brief is where the expensive decisions happen, and it should fit on one page. Working title, reader, the problem they have, the promise you make, the central claim, the shape of the evidence, the length, and the subjects this piece will deliberately leave alone.
Non-goals do more work than anything else on that page. Long-form writing expands until something stops it, and the cheapest stopper is a line written before you became invested in the tangent. The test here is agreement: you and your approver have read the same brief.
A stage that ends without a decision is just a place where work waits.
Stage three: research to a stopping rule
Research without a stopping rule expands to fill whatever time remains. Before you begin, list the questions the piece must answer. Research is over when each question has an answer, a named source, or an explicit note saying you could not resolve it.
Log as you go, in your own words, with the source attached. Copying passages verbatim into a research file creates a slow problem: months later you cannot tell your summary from someone else's sentence, and the only safe response is to check every line again.
Expect gaps. A named gap is a normal research output rather than a failure. Assign each one to a follow-up call, to a caveat in the finished text, or to a decision to cut the section that depended on it entirely.
Stage four: outline as a sequence of promises
Write the outline in full sentences, one per section, each stating what the reader will understand by the end of it. Fragments such as background hide the fact that nothing has been decided yet, and they let a weak section survive until the draft is too heavy to restructure.
Then read the sentences in order without the headings. If the sequence argues the case on its own, the structure holds. If it reads as a list of topics, you have a collection, and reordering now costs minutes instead of days.
Stage five: draft in passes, not in polish
Draft to reach the end, not to reach quality. The first complete pass exists to show you the real shape of the argument, which is never quite the shape the outline predicted. Start with the section you understand best; openings are much easier once the piece exists.
Work section by section with only that section's research in front of you. The whole file open at once is a reliable way to write paragraphs that mention everything and land nothing. Leave visible markers wherever you stop: an unanswered question, a missing figure, a transition you have not solved.
Stage six: revise from structure downward
Run separate passes and name each one. Structure first: is anything out of order, repeated, or absent. Then evidence: does every load-bearing claim rest on a source you have actually read. Then clarity, then voice, then length. Leave the draft alone for a night between passes if the deadline allows.
A single pass trying to fix everything blurs priorities. You will smooth a paragraph you should have deleted, and the structural problem will still sit there underneath better sentences. Named passes are easier to finish, because each of them has an obvious end.
The finish line and what comes after
Define finished before you are near it. For most long-form work, finished means the argument is complete, every claim is sourced, the piece does what the brief promised, and at least one other person has read it from beginning to end without stopping.
Then record what you learned while it is still fresh. Which stage took longest, which research you wish you had done first, which section you rewrote three times. Two paragraphs after each piece shape the next one more than any general workflow advice.
One idea, with the decision recorded at every stage
Backlog entry: “Piece about remote hiring. Interesting angle. Maybe 2,000 words?” Assigned a publication date the same afternoon.
Idea — Reader: engineering managers making a first remote hire. Question: how do I judge someone I will never meet? Promise: a defensible process, not reassurance. Cleared 3 March. Brief — Claim: structured work samples predict remote performance better than interviews. Non-goals: visas, salary bands, tooling. Accepted 5 March. Research — Six questions, five answered with named sources, one carried as a caveat. 12 March. Outline — Five section sentences read in order and argue the case. 13 March. Draft — Complete end to end, four markers left. 19 March. Revision — Structure, evidence, clarity, voice. Read by an editor. 24 March.
None of those lines is a status update. Each records something that had to be true before the next stage began, which is why the piece never needed a rescue week.
Common mistakes
| The mistake | Why it costs you, and what to do instead |
|---|---|
| Scheduling an idea before qualifying it | A publication date makes an idea feel real, so it lands on the calendar while the reader and the promise are undefined. The deadline then forces a shape onto material that never fitted it. Schedule only what survives qualification. |
| Researching until you feel ready | Readiness is a feeling, and it arrives late or never. Without a stopping rule you keep reading, because reading is more comfortable than committing to a claim. Write the questions first and close them one by one. |
| Editing while drafting | Fixing the sentence you have just written is the most tempting interruption in writing, and it hides structural trouble behind attractive prose. Draft to the end, mark the problems as you pass them, and revise separately. |
| Treating publication as the end | Once the piece ships, the workflow feels complete and nothing gets written down. The next project then repeats the same delay at the same stage. Spend ten minutes recording what slowed you and what you would sequence differently next time. |
Your checklist
- Write the reader, their question, and your answer before the idea gets a date.
- Keep the brief to one page and include at least three non-goals.
- List the questions research must answer, then stop when each is closed.
- Turn the outline into one full sentence per section and read them in sequence.
- Draft to the end, leaving markers rather than stopping to repair.
- Run structure, evidence, clarity, and voice as separate passes.
Key terms
- Stage gate
- A checkable condition that must be true before work moves on, such as an accepted brief or a research list where every question has an answer or a named gap.
- Stopping rule
- A rule agreed before research begins that defines when there is enough material, usually a fixed list of questions rather than a feeling of being ready.
- Revision pass
- One trip through a draft looking for a single class of problem, such as structure or evidence, instead of attempting to repair every kind of weakness at once.
Setting a project up properly?
Frequently asked questions
How long should each stage take?
+
There is no fixed ratio, but the early stages are usually underweighted. For a 3,000-word piece, spending a full working day on the brief and the outline is normal, and it tends to return days that would otherwise go into restructuring.
Can stages overlap?
+
Research and drafting overlap comfortably. Brief and outline should not. Deciding the structure while the promise is still moving produces an outline you have to rebuild twice. Keep the test before drafting firm even when everything else around it stays fluid and unfinished.
What if the brief turns out to be wrong?
+
Go back, rewrite it, and re-outline. That feels like losing ground and is far cheaper than carrying a stale promise all the way through revision. Note what changed and why, so the same misjudgment is easier to spot the next time it appears.
Does this work for a book-length project?
+
The sequence holds, applied twice: once for the manuscript as a whole and once per chapter. The brief grows longer, research runs continuously, and the tests matter considerably more, because a wrong structural decision now costs weeks of writing rather than a single afternoon.
Where does Strut AI fit into this workflow?
+
Best at the edges of a stage. Ask the assistant to list questions your brief has not answered, to propose three structures for an outline you wrote, or to find claims with no source attached. The decisions stay with you.
How do I stop the workflow becoming bureaucracy?
+
Judge each test by whether skipping it has ever hurt you. If a check has caught nothing across ten pieces, delete it. Four tests you genuinely apply are worth more than eight you route around when the deadline gets close.