✍️ AI Killed the Hard Part of Writing. Now the Harder Part Is All That's Left.

Jun 15 • From AIA

For most of the people we talk to, writing used to be the bottleneck. The blank page, the slow start, the draft that took two hours to get to a point where it felt workable. That bottleneck is largely gone now. A capable AI model can produce a usable first draft in under two minutes. The hard part of writing: getting words on the page, has become nearly effortless.

What nobody warned us about is what happens next. When first drafts are cheap, editing becomes the job. And most people's editing process was designed for a world where drafts were expensive and rare, not fast and abundant.

The result is a growing backlog of AI-generated content that's good enough to feel like it almost works, but not quite good enough to use without significant revision. A growing awareness is setting in that the revision is taking longer than the writing used to.

------------- Context -------------

The economics of writing have flipped. Before AI, time was heavily front-loaded. Research, outlining, drafting: these consumed the majority of hours, with editing as a finishing step. A piece of content that took three hours might have involved two and a half hours of creation and thirty minutes of editing.

Now the ratio has inverted. A draft that takes two minutes to generate might need forty-five minutes of editing to reach a standard worth publishing. The total time is still less than before, but the distribution has changed, and the nature of the work has changed with it.

Editing is harder than drafting in one important respect: it requires holding the standard for quality in your head while simultaneously evaluating whether what's in front of you meets it. Drafting lets you externalize thinking. Editing requires you to internalize a clear picture of what good looks like and apply it consistently to every sentence, paragraph, and argument in the piece.

Most people haven't developed that capacity deliberately, because most people haven't needed to. The drafting process used to do a lot of the thinking work. The act of writing was also the act of figuring out what you were trying to say. AI drafting removes that process, which means the thinking has to happen somewhere else, usually in the editing phase, which is why AI-assisted editing often takes longer than it seems like it should.

------------- When the Volume Increases, the Editing Problem Compounds -------------

The obvious response to cheap drafts is to produce more of them. More content, more proposals, more client communications, more social posts. The production ceiling goes up, so the volume expands to fill it.

What expands with it is the editing queue. And an editing queue is different from a drafting queue in one important way: it doesn't wait patiently. Unedited content is unfinished work, and unfinished work creates mental overhead. It's the low-grade background awareness that something needs attention. A growing editing backlog generates a specific kind of stress that a growing drafting queue never did, because with drafting, you at least know the work hasn't started yet.

A content creator who adopted AI for her newsletter production discovered this dynamic six months in. She'd gone from one newsletter per week to three, because drafting was now so fast. But she was spending more total time on the newsletter work than before, because each piece required thirty to forty minutes of editing, and she was producing three times as many pieces. Her total writing time had actually increased. The drafts were faster. The output was slower.

The bottleneck hadn't been eliminated. It had been relocated.

------------- Editing Is a Skill, and It Can Be Developed -------------

The good news is that editing skill responds to deliberate practice in the same way any skill does. And unlike drafting, which AI has largely taken over, editing is a deeply human capability. It requires judgment, taste, and the ability to hold context across an entire piece while evaluating individual decisions. AI can assist with editing, but the standard itself has to come from somewhere, and that somewhere is the person doing the work.

The professionals who are managing the new economics of writing well have developed a clear editorial standard: an explicit, documented picture of what good output looks like for their specific context. Not just "it should sound like me," but specific criteria. What makes an argument clear, what counts as a strong opening, what level of specificity is right for this audience, what common patterns need to be caught and fixed.

With that standard documented, editing becomes faster and more consistent, because the judgment work has been done in advance rather than reconstructed on every piece. A freelance strategist who built a detailed editorial checklist for her client proposals found that her editing time dropped from forty minutes per proposal to about twenty. Not because the proposals got easier to edit, but because she stopped re-deciding what good looked like every time she opened a draft.

That two hours per week recovered across a fifty-week year is a hundred hours. Not from better AI. From a clearer editing process.

------------- Practical Moves -------------

First, build an explicit quality standard for your most common content types. Not a vague aspiration but a specific checklist: what does a strong opening look like, what level of evidence is required, what makes a conclusion land well. This standard does the judgment work once instead of on every draft.

Second, separate editing sessions from generation sessions rather than moving immediately from one to the other. The mental mode required for each is different, and switching between them in the same sitting produces slower, lower-quality editing.

Third, give AI a role in the editing process too. Not as the arbiter of quality, but as a first pass for surface issues. Use it to catch inconsistencies, redundancies, and structural problems before you do the judgment-level editing yourself. This pre-clears the queue so your editing time goes toward the decisions that actually require your judgment.

Fourth, set a time budget for editing before you start and hold it. Open-ended editing sessions expand to fill available time. A firm limit forces prioritization and prevents the diminishing-returns revision cycles that consume hours without meaningfully improving output.

Fifth, track your editing time per content type for two weeks. The number is almost always surprising, and seeing it clearly is what creates the motivation to build a better process around it.

------------- Reflection -------------

AI hasn't made writing easier in the full sense of the word. It's made drafting easier, which is genuinely valuable, while making editing more important, more frequent, and more demanding. That shift is permanent, and it rewards people who treat editing as a skill worth developing rather than just a step worth getting through.

The time that AI returned through faster drafting is available to be reinvested. The question is whether it gets reinvested in more drafts, or in the editing capability that makes every draft worth the time it takes to finish.

What does your current editing process actually look like?

If you're honest about the time it costs, is it keeping pace with what AI is producing?

6 comments

✍️ AI Killed the Hard Part of Writing. Now the Harder Part Is All That's Left.