Key takeaways

Don't ask AI to 'write a blog post'. Ask it to do specific, mechanical sub-tasks inside a workflow you control.
Outline first, draft second. The outline is where the thinking happens. The drafting is where AI helps most.
Always do the final pass yourself. AI will smooth out your voice if you let it.
Save the prompts as templates. The payoff builds when you stop retyping them.
Track your before and after time. The gains are real but smaller than the hype claims.

A piece on the consolidation of AI image tools, 1,400 words, took me 68 minutes to write. A post of the same length on a similar topic, written the month before without help, had taken 3 hours 40 minutes by my time tracker. The faster one was slightly better.

I hadn’t asked Claude to write it. Type “write a 1,200-word post about X” into one of the AI writing tools and you get something grammatical and bad. Generic structure, stock transitions, no argument, and a skeleton too wrong to fix. What worked was four narrow prompts at four moments, with me doing the parts that need judgement.

0 min125 min250 min375 min500 minOutline + ClaudeAll-human (prior month)
One comparable pair: with Claude vs without 68 minutes vs 3h 40min (220 min) for two ~1,400-word posts of similar length and topic, by the author's own time-tracker. It's one pair of posts, not an average.
One comparable pair: with Claude vs without
CategoryMinutes
Outline + Claude 68 min
All-human (prior month) 220 min
Cite or embed this

That’s one pair of posts, and I haven’t timed it across many. Pieces built on reporting should gain less, because the writing isn’t the slow part. So should pieces whose value is a strong voice.

The four steps took about 10 minutes to find the angle, 15 to outline, 30 to draft and 15 to polish.

1. Argue about the angle

The angle is the hardest part of any piece. Most AI workflows skip it, which is why their posts are generic.

I'm writing about [topic] for an audience of [audience]. My initial
take is [your rough thesis].

Before I outline, push back. What's the most interesting *contrarian*
angle in this space right now? What's the angle most other writers
are missing? What would surprise this audience?

Give me 5 angles I could take. For each: one sentence on the angle,
one on why it's surprising, one on the strongest counterargument.
Don't be polite: be useful.

Without “be useful”, the model opens with “great question” and offers five angles with no edge. Spend ten minutes, choose an angle (yours or one it suggested) and move on.

2. Write the outline yourself, then have it attacked

The argument lives in the structure. Hand over the outline and you’ve handed over the thinking. Write it as a flat list of bullets, then ask for a critique.

Here's my outline for a piece on [topic, angle]:

[paste outline]

Critique it before I draft:
1. Is there a section that's actually two sections?
2. Is there a section that's redundant?
3. What am I missing that a smart reader will want?
4. Is the structure burying my best argument? What should I lead with?
5. Suggest one section I haven't included that would strengthen the piece.

Be specific. Reference the bullets by number.

On my image-tools piece, this step caught a structural problem I’d missed.

3. Draft one section at a time, then rewrite it

Give Claude one section’s bullets at a time, never the whole post. Whole-post drafts lose track of the argument and go bland.

Below is the outline for the full piece, then the bullets for the
section I want you to draft now.

FULL OUTLINE (for context, do not draft):
[paste]

DRAFT THIS SECTION ONLY:
[paste this section's bullets]

Tone: [conversational/analytical/etc.]. Length: ~250 words.

Voice notes: I write short paragraphs, use specific examples, and
never use the phrases "in today's fast-paced world," "delve into,"
"unlock the potential of," or "navigate the landscape of."

If you need to make up a stat or example, use [BRACKETS] so I can
see it and replace it.

Claude will sometimes invent statistics or examples that sound right. The brackets force it to show you what it made up. Each section then takes four or five minutes of rewriting: restore your voice, replace the brackets with real examples and tighten.

4. Ask for ten headlines, then choose one yourself

AI is good at producing ten of something and bad at picking the best. Use it for range and keep the choice.

Here's my piece:

[paste full draft]

Generate:
1. Ten headlines. Range across: blunt-statement, contrarian, list-format,
   question-format, character-driven. Don't repeat any phrase across the ten.
2. Five alternate opening sentences for the piece. Each should immediately
   commit to a different angle.
3. Three options for the closing line.

For each, no explanation: just the variants.

Sometimes it finds a headline I’d never have written. Sometimes I stitch one together from three.

It doesn’t burn out your taste, because you’re still doing the parts that exercise it.

What to keep for yourself

Reporting. If a piece’s value is in its reporting, don’t shortcut the interviews.
A strong voice. If your voice is distinctive (funny, angry or very particular), AI drafts will sand it down. Use only the outline critique and the headline list.
Anything with real stakes. An important email, sensitive client work, anything legal: write it yourself from start to finish.

For the rest of everyday work writing, save the four prompts as templates so you never retype them.

Three-minute lesson on this Build a weekly research brief Marketing course, advanced. Free, no sign-up.

More from the blog

All articles →