1. Start with the job, not the tools

Every writing system that fails starts the same way: someone picks a shiny app, then tries to fit their work into it. Start from the other end. Write down the four jobs your system has to do: catch ideas, turn them into drafts, edit the drafts, and publish on a rhythm you can keep.

That’s it. Four jobs. If a tool doesn’t clearly serve one of them, it doesn’t belong in your system yet — no matter how many people recommend it. I spent a month once trying to run my whole workflow inside a note-taking app that was brilliant at notes and terrible at drafts. The app wasn’t the problem. I’d never actually listed the jobs.

Do this on paper, in ten minutes. Four headings, one sentence each about what “done” looks like. That page is your system spec, and it will save you from every tempting tool that solves a problem you don’t have.

2. Capture everything in one place

Ideas are cheap and forgettable in equal measure. The fix isn’t a better memory — it’s a single inbox where every idea goes the moment it appears, no matter where you are.

Mine is one note called “Ideas,” and the rule is absolute: an idea that isn’t in the inbox doesn’t exist. A half-formed headline on the train, a reader’s question in an email, something a colleague said in a meeting — it all goes in, in whatever messy words it arrived in. I don’t judge ideas at capture time. Judging is a different job, and doing it at the wrong moment is how good ideas die.

Once a week I spend twenty minutes sorting the inbox. Most ideas get deleted. A few get a one-line note about what the piece would actually say. Two or three graduate to the draft queue. That weekly sort is the most valuable twenty minutes in my whole system, because it’s the moment I choose what to write instead of staring at a blank page wondering.

3. Give every draft the same first pass

The blank page is where writing systems go to die. So don’t start there. Give every draft the same first pass: a short outline, written by you, before any AI tool touches it.

My outline is always the same shape — the point of the piece in one sentence, three or four headings, and one concrete example per section. It takes fifteen minutes and it does two things the AI can’t do for you: it forces you to know what you’re actually saying, and it gives the draft a spine that survives editing.

Only then do I ask the AI to draft from the outline. The model gets structure and examples to work with, and I get a first draft that’s mine in shape even when the sentences aren’t. Without the outline, the AI picks the structure — and its structure is always the same safe, generic one. With it, the piece starts out yours.

4. Keep a voice file the AI can’t overwrite

Here’s the part most people skip, and it’s the part that keeps the system from turning your writing into mush. Keep a voice file: a short document that describes how you actually write, with examples.

Mine is about a page long. It lists the phrases I never want to see (“in today’s fast-paced world,” “delve,” “tapestry”), the sentence length I aim for, and three paragraphs I’ve written that sound like me on a good day. Every time I brief an AI tool, that file goes in with the prompt. Every time I edit, I check the draft against it.

Update it when you notice something. If a reader tells you a piece sounded like you, save the sentence that did it. If you catch yourself deleting the same AI tic for the tenth time, add it to the banned list. The voice file is the one part of the system that compounds — a year from now it will be the most valuable document you own.

5. Set a weekly rhythm you can actually keep

A system you can’t keep on a bad week isn’t a system, it’s a fantasy. So design the rhythm for your worst week, not your best one.

Mine is embarrassingly small: one idea sorted, one draft outlined, one piece edited, one piece published — per week. Some weeks I do more, and that’s a bonus, not the plan. The plan is the minimum that keeps the machine moving. A single published piece a week beats four pieces in a heroic weekend followed by a month of silence, because readers and search engines both reward consistency over bursts.

Pick a day for each job and protect it. Monday is my sort day, Wednesday is drafting, Friday is editing and publishing. If a week goes sideways, I protect Friday first — the publish day is the heartbeat of the whole system.

6. Build a small library of prompts that work

Stop writing a fresh prompt every time you open an AI tool. The moment a prompt produces a draft you like, save it — the exact wording, not a summary of it.

I keep about a dozen. One for turning an outline into a first draft. One for tightening a flabby paragraph. One for checking a draft against my voice file. One for generating five headline options that don’t all sound the same. Each one took a few tries to get right, and each one now saves me ten minutes and a blank prompt box every single time I use it.

Treat them like recipes: test them, tweak them, and throw out the ones that stop working. A prompt library is the closest thing to leverage in this whole system — the same good thinking, reused, instead of reinvented at 9 a.m. on a Tuesday.

7. Review the system once a month — and prune it

Systems rot. A tool stops being worth its subscription. A habit that helped in January is dead weight by June. So once a month, spend thirty minutes asking three questions: what did I actually use, what did I skip every time, and what’s the one friction point that slowed me down most?

Last year that review killed two subscriptions and one elaborate tagging scheme I’d built and never once used. It also surfaced the fact that my editing step was taking twice as long as it should have — which is how the checklist I now run on every draft was born. The review is where the system gets better. Without it, you end up maintaining a museum of good intentions.

Prune ruthlessly. A system with four working parts beats a system with twelve parts where eight are decoration. Every piece you remove is a decision you no longer have to make.

The honest truth

A personal writing system isn’t a productivity hack. It’s a promise to your future self that the next piece won’t start from nothing. The AI tools are the fast part — they draft in seconds and never get tired. But speed was never the bottleneck. The bottleneck was deciding what to write, starting the draft, and finishing it in your own voice. That’s what the system is for.

Build it small, build it boring, and keep it for a month before you judge it. The writers who publish consistently aren’t more talented or more disciplined than you. They just stopped rebuilding the road every time they wanted to drive somewhere.

Frequently Asked Questions