AI Writes My First Draft, Not My Final Draft (And That's the Point)

The blank page used to cost me an hour before I wrote a single word I liked. I'd open a blank document, stare at the cursor, write a sentence, delete it, write another, delete that one too. An hour in, I'd have a paragraph I still didn't love and a full day of other work waiting behind it. This is hack 3 in my "30 Hacks From Dahlia" series, thirty specific things I actually do that save real time, and this is the one that changed how I write everything: emails, blog posts, proposals, even the page you're reading right now.

Why starting from blank is so expensive

An hour sounds small until you count how often it happens. Every email that needs more than two sentences, every proposal, every blog post, every caption that has to say something real, all of it used to start the same way: an empty screen, no structure, no momentum. The cost isn't only the hour. It's the decision fatigue that piles up before you've written a single word, and the fact that by the time you finally have something on the page, you're already tired of the piece you still have to finish.

The fix isn't writing faster. It's never starting from zero. Once there's something on the page, even something rough and wrong in places, my brain switches from generating to editing. Editing is faster, and for me, editing is where the actual writing happens anyway.

What I actually feed it (not a polished brief)

The instinct is to write AI a clean, organized brief before asking for a draft. I stopped doing that. A clean brief takes almost as long to write as the piece itself, and putting it together tends to strip out the specific, half-formed details that make a draft usable. Instead I feed it whatever I already have lying around: rough notes I typed out in two minutes, a voice memo I recorded in the car, a list of bullet points that don't connect to each other yet, half a sentence I texted myself the night before. Messy input is fine. It's better than fine, because the mess is where my actual thinking already lives.

This matters more than it sounds like it should. A polished brief has already been filtered through "does this sound smart," which is exactly the filter that flattens your voice before the AI even sees it. Rough notes haven't been filtered yet. They're closer to how you actually think, which means the draft that comes back is closer to how you actually sound, even in its messy, unfinished state.

The prompt structure that gets a useful messy draft

The prompt matters more than the notes do. If I ask for a finished piece, I get something generic that reads like it was written for everyone, which means it doesn't sound like me specifically. What I ask for instead is a rough first draft built only from what I gave it, told explicitly not to invent facts or stories I didn't include, and told explicitly that it's a starting point I plan to rewrite.

Here are my rough notes on [topic]. Write me a messy first draft, not a finished piece. Use only what I gave you, don't invent examples, stats, or stories I didn't mention. Keep the structure loose, I'll rewrite the sentences myself. I just need something to react to.

That last line does most of the work. Telling it you plan to rewrite the sentences changes what comes back: less polish, less padding, more of a scaffold you can actually see through to the structure underneath.

The human pass: cut, add the real story, fix the voice

Once the draft exists, my job starts, in this order. First I cut. Most first drafts, AI-written or not, run long and repeat themselves, so I cut whole sentences and sometimes whole paragraphs before I touch a single word choice. Second I add. This is the part AI can't do for me: the specific story, the detail from an actual conversation, the line that only makes sense because I was really there. Third I fix the voice, sentence by sentence, until it sounds like something I'd genuinely say out loud, not something that merely reads correctly.

Why the 60 percent framing matters

The number that matters here isn't the two minutes. It's the 60 percent. A messy first draft gets me roughly two thirds of the way to something usable, a structure, a rough argument, the general shape of the piece, before I've spent any real energy at all. That means the energy I do spend goes entirely toward the remaining 40 percent: the real story, the specific detail, the sentence that couldn't have come from anywhere but my own experience. I'm not skipping the writing. I'm skipping the part of writing that was never where the value lived in the first place, the blank page.

The magic isn't that it writes for me. It's that it gets me to 60 percent in two minutes, so I spend my energy on the part that sounds like me.

The two mistakes that undo this

The first mistake is publishing the AI draft close to as-is. It reads fine on a first pass, which is exactly the trap: fine isn't the same as sounding like you. Skip the cut-add-fix pass and you end up with content that's technically correct and completely forgettable, and readers can usually tell even when they can't say exactly why.

The second mistake runs the opposite direction: treating every sentence like it deserves careful editing, instead of making the big cuts first. If you polish the word choice on three paragraphs you're eventually going to delete anyway, you've spent your limited editing energy in the wrong order. Cut structurally first. Fix voice only on what's actually going to stay.

This is hack 3 of 30. The rest of the series covers the other small, specific things I actually do that quietly save time without cutting corners on the parts that matter. If the blank page has been costing you an hour too, this is the one hack I'd start with.

Dahlia Imanbay

AI Strategist, Fractional CMO, and Full-Stack Developer with 16+ years of experience building AI systems for healthcare, SaaS, and mission-driven brands. Writes from production experience, not theory.