The Minto Pyramid: Make Claude Write Like a Consultant
Claude buries the point by default. The Minto Pyramid flips it: answer first, three reasons, then the evidence. Here are the 5 prompts, the Project that applies it on its own, and the style file that keeps your voice.
August 30, 2026 · 8 min read
By default, Claude buries the point. It warms up, sets context, and drops the conclusion at the very bottom, where nobody reads. Your reader, meanwhile, decides in the first sentence whether to keep going. The Minto Pyramid flips the order: the answer first, three reasons next, the evidence last.
Barbara Minto formalized this principle at McKinsey, and the firm still trains its consultants on it. It is why a consulting memo reads in one pass. This guide installs that structure into Claude: the 5 prompts to restructure any draft, the Project that applies it without you asking again, and the style file that keeps your voice.
The Claude AI Lab is my Skool community where I share my Claude systems and the more advanced modules. Access is $67/month.
Join the Lab →The pyramid, actually
The principle sits in three layers, and the order is everything.
No overlap between them, and nothing important left outside them. If a fourth reason is fighting to get in, it is almost always evidence in disguise: it goes under one of the three, not beside them.
Here is what the flip looks like, on a real team update. Before, the way most people write it:
Hi team, following up on last week's session. We looked at the Q3
numbers and there were a few things worth flagging. Onboarding time
has been creeping up, and we also saw some churn in the smaller
accounts. Support tickets are up too, though some of that is
seasonal. Given all of this, I think we should probably consider
moving the onboarding rebuild forward.
After, once run through the prompts, the same information becomes:
We should move the onboarding rebuild to Q4 and delay the reporting
work.
1. Onboarding time is now our largest churn driver. Accounts taking
over 14 days to onboard churn at roughly three times the rate of
those under 7.
2. The fix is smaller than we assumed. Two of the four blockers are
template work, not engineering.
3. Delaying reporting costs us little. No customer has asked for it
this quarter and no renewal depends on it.
Same information. The difference is that the second version can be acted on after one sentence, and every line after it defends that sentence instead of building up to it.
The 5 prompts
Five prompts, each with a precise job. One point of order before you start: run prompt 2 BEFORE prompt 1 on anything you have already written, because you need to know what your draft was secretly arguing before you restructure it.
Prompt 1, the restructure.
Rewrite this using the Minto Pyramid: open with my single main
conclusion in one sentence, then exactly three supporting reasons,
then the evidence under each. Cut anything that does not sit under
one of the three.
Prompt 2, the buried lede test.
Read this draft and tell me the one conclusion it is actually
building toward. If I only had one sentence to give my boss, what
would it be? Then tell me where in the draft I hid it.
Prompt 3, the three-reason audit.
Check my three reasons. Do they overlap, and together do they fully
cover the case? Flag any that is a detail pretending to be a reason.
Prompt 4, the so-what pass.
Go line by line and ask so what after each one. Delete every line
that cannot answer it. Show me what survives.
Prompt 5, the email version.
Compress this into an email under 150 words in the same structure:
answer, three reasons, evidence only where it earns its place.
Keep my voice.
Running prompt 1 on a draft where you have not decided what you think yet. The pyramid organizes a conclusion, it does not create one. If you cannot answer prompt 2 yourself, the problem is your thinking, not your formatting.
The Project that structures by default
One Project, and you stop pasting rules. Everything you draft inside it comes out in the structure, without asking. Open Projects in Claude, create one, name it "Pyramid Writer", and paste the block below into the Project instructions. Then add the style file from the next section to the Project knowledge.
You write in the Minto Pyramid structure by default, for every draft,
without being asked.
Structure rules:
1. Open with one sentence carrying the single main conclusion, ask,
or recommendation.
2. Follow with exactly three supporting reasons. They must not
overlap, and together they must cover the whole case.
3. Put evidence under the reason it supports. Never above the
conclusion.
4. Anything that does not sit under one of the three reasons gets
cut, not relocated.
Behaviour rules:
- If I have not made my conclusion clear, ask me for it before
drafting. Do not guess a conclusion and build a pyramid on it.
- Never open with background, context setting, or a warm up sentence.
- Match the voice in the style file. Do not write in consulting jargon.
- After every draft, show me in one line what the three reasons are,
so I can check them fast.
- Default to short. Only go long when I ask.
Dumping these rules into a plain chat instead of the Project is the mistake. The chat forgets, the Project does not. That difference is the whole point: you paste a messy draft with no instructions at all, and the answer comes back on top.
The style file
Structure without voice reads like a template. The style file stops that. You make it once, add it to the Project knowledge, and Claude reads it on every task. Create a document writing_style.md with these five sections:
# WRITING STYLE
## Who I write to
Who reads my drafts, what they care about, and how much detail
they want. Name the two or three recurring readers.
## My voice in three rules
Three specific rules. Not adjectives. For example:
- Short sentences. One idea each.
- No jargon a new starter would not understand.
- I ask for the decision directly, I do not hint at it.
## Words I use
Paste 5 to 10 phrases you actually write. This is what Claude
matches against.
## Words I never use
The corporate filler you hate. Circle back, leverage, synergy,
reach out, touch base, per my last email.
## Two examples
Paste one email you were proud of and one that got a fast reply.
Label them GOOD. These do more work than every rule above.
Claude matches patterns far better than it follows adjectives. Two real emails beat a paragraph describing your tone. The mistake to avoid: describing your voice as "professional but friendly". That describes every writer alive and tells Claude nothing. Rules and examples only.
The weekly loop
The prompts fix a draft. The loop fixes your writing. Once a week, take the three longest things you sent and run this:
Here are three things I sent this week: [paste].
For each one:
1. Tell me where I buried the conclusion and how many sentences the
reader had to get through before reaching it.
2. Name the pattern I keep repeating across all three.
3. Give me one rule to add to my style file that would have caught it.
Add whatever rule it gives you to the style file. After four or five weeks the style file does the work and you reach for prompt 1 less and less often.
The mistake to avoid: running the loop on your best writing. Give it the rushed ones, the Friday-evening ones. The pattern lives there. You will know it worked when the loop starts telling you it found nothing worth flagging.
The pyramid comes down to one discipline: the answer on top, and every line below it defending that answer instead of building up to it. The prompts set it on a draft, the Project makes it permanent, the style file keeps it in your voice.
Take the last long email you sent and run prompt 2 on it right now. It takes ninety seconds and it shows you exactly where you hid the point. Everything else is just making sure you never hide it again.
Want to go further?
In the Lab, I share my prompts, templates and how I chain them to ship real work, not fluff.
A dedicated session or program, tailored to your tools and use cases.
And day-to-day, I post one reel a day on Instagram: @quentin_iamarketing