Build Leaner  /  50 AI prompts

Free download · 31-page PDF

50 AI prompts you'll
actually use every day.

The everyday prompts almost everyone writes badly — email, meetings, writing, decisions, spreadsheets, client work — rewritten as working templates. Every one is explained below, and the whole set is a free PDF you can keep.

50prompts
10everyday jobs
1formula behind them
0cost
Free · 31-page PDF

Get the full pack

Enter your email and the PDF downloads on the next screen. A copy lands in your inbox too.

No spam. Unsubscribe in one click. You'll also get the weekly build.

The diagnosis

Why most AI prompts come back useless

It is almost never the model. When ChatGPT, Claude or Gemini gives you something bland, hedged and unusable, the cause is nearly always one of these five — and usually more than one at a time. Every prompt in this pack is built to avoid all five by default.

01

You asked a generalist

With no role or reader named, you get the average of everything ever written on the topic.

02

You gave it nothing to work with

The model fills every gap you leave with the most average option available. Generic in, generic out.

03

You didn't say what it should look like

Ask for a table and you get a table. Say nothing and you get six paragraphs to read through.

04

You never defined "good"

"Make it better" has no target. "Cut 30%, keep every number, remove all hedging" does.

05

You accepted the first draft

The first output is a starting position. One round of "now critique that and improve it" beats a perfect opening prompt.

The formula

The five blocks behind every prompt

You don't need to memorise fifty prompts. You need the structure they share. Once you can see these five blocks, you can write your own for any task — which is genuinely the point of the pack. Short on time? Blocks two, three and four get you about 80% of the way there.

1WHO

Who's answering, and who the answer is for.

2CONTEXT

The real material — your notes, your email, your numbers.

3TASK

One clear job, verb first.

4FORMAT

Exactly what the output should look like.

5THE BAR

What separates a good answer from a passable one.

All 50, explained

The prompts, and what each one is for

Ten sections, five prompts each, covering the jobs that actually fill a working week. Below is what each prompt does and why it's built the way it is. The full copy-paste templates — with the blanks marked and the reasoning spelled out — are in the free PDF.

01

Email & inbox

Email is where most people lose their first hour of the day. These five turn a queue into a checklist you can clear in one pass.

Prompt 01

The reply that ends the thread

Instead of asking AI to invent your position, you give it your rough unpolished note and let it edit. It comes back under 120 words, with one clear ask and a dated next step — which is why the thread actually ends instead of bouncing back and forth for three more messages.

Prompt 02

Inbox triage in one pass

Paste a dozen senders and subject lines and get one table back: summary, action, effort, urgency. The critical instruction is that it must not draft anything yet. Sorting and writing are different jobs, and mixing them is why inbox sessions sprawl.

Prompt 03

Say no and keep the relationship

Most declines fail because they hedge, leaving the other person to hear a maybe. This one forces the refusal into a single unambiguous sentence, gives one honest reason, offers an alternative where you have one, and caps the apologising at once.

Prompt 04

The follow-up that gets answered

Chase-ups die when they're about you. This rewrites the ask as a yes/no question the recipient can answer in five seconds and gives them a reason it matters on their side. You get three temperatures back — friendly, neutral, final nudge — so you never rewrite.

Prompt 05

Decode a difficult message

Separates what someone literally asked for from what they probably want underneath, rates the emotional temperature, flags what's ambiguous, and gives you two possible replies: one that concedes, one that holds your ground. Useful before you answer anything tense.

02

Meetings & notes

A meeting only pays for itself if the decisions survive it. These five make the notes do the work instead of sitting in a doc nobody reopens.

Prompt 06

Messy notes to decision log

Turns raw scribbles into four clean blocks: decisions, actions with owners and dates, open questions, and risks. The guardrail is the whole point — it must write OWNER UNKNOWN rather than guess, so you never end up circulating a confident, fictional plan.

Prompt 07

Meeting prep in five minutes

Gives you three target outcomes (best, realistic, minimum acceptable), the five questions to ask, the three objections you'll hear with answers, and your opening sentence. Naming the minimum acceptable outcome before you walk in is how you stop conceding in the room.

Prompt 08

An agenda that saves the meeting

Every item has to name the decision it exists to make, which quietly kills all the status-update items. It also produces a "not in this meeting" list for the topics people will try to drag in, and it budgets the minutes so you finish on time.

Prompt 09

The recap nobody argues with

A neutral, under-200-word summary split into decided, next, and still open — ending with a correction deadline. If nobody objects by the date, your version becomes the official record. That single line is what makes recaps worth writing.

Prompt 10

Should this even be a meeting?

An honest read on whether your calendar invite could be an email, a commented doc, or a ten-minute call — and then it drafts whichever cheaper version it recommended. You leave with the thing itself, not just an opinion about your calendar.

03

Writing & editing

AI is a mediocre writer and an excellent editor. These five use it in the direction it's actually good at.

Prompt 11

Shorter without worse

Cuts a piece to a set percentage while keeping every fact, number and commitment, stripping only hedging and repetition. Then it lists what it removed and why — so a lossy edit becomes a reviewable one and you can put back anything that mattered.

Prompt 12

Write in my voice

The reliable way to get AI to sound like you isn't adjectives, it's samples. You paste three things you've written, it produces a description of your voice — sentence length, rhythm, punctuation habits, what you never do — then writes using it. Save that description and reuse it.

Prompt 13

The brutal edit

A skeptical editor that tags your draft [CUT], [VAGUE], [UNPROVEN] and [BURIED], then names the single biggest weakness in one paragraph. Crucially it doesn't rewrite anything, so you stay the author rather than becoming the approver of someone else's draft.

Prompt 14

Rewrite it for the approver

Reframes a document around one specific reader and their time budget: conclusion first, the ask in the opening two sentences, everything else pushed below a Detail line. Describing who reads it does far more than any instruction about tone.

Prompt 15

Ten headlines, then the argument

Ten titles across five different angles — number-led, question, contrarian, outcome-led, plain. Then it ranks its top three, argues for each, commits to one, and tells you what's risky about it. The reasoning is usually more useful than the list.

04

Thinking & deciding

The highest-leverage use of AI, and the one almost nobody reaches for. Five prompts for when the answer matters more than the wording.

Prompt 16

The decision teardown

Three real options including one you probably haven't considered, the strongest case for each, and — the key part — what must be TRUE for each to be right. That converts a vague choice into a list of things you can go and verify. Ends with a recommendation and a confidence percentage.

Prompt 17

Argue against me

AI defaults to agreeing with you, which makes it useless as a sounding board. This asks for the strongest case a smart person who genuinely disagrees would make, then tells you which objections actually matter and what cheap test would settle it within a fortnight.

Prompt 18

The pre-mortem

You assume the project has already failed six months from now and ask for the post-mortem. Assuming failure has happened gets you specific, concrete causes instead of polite risk categories — plus the warning sign you'd have seen in week two.

Prompt 19

Find my blind spot

You paste your actual reasoning and it flags assumptions you're treating as facts, small-sample generalisations, and places you might be solving the wrong problem — quoting your own words back so the critique is checkable rather than abstract.

Prompt 20

The 10/10/10 gut check

How will you feel about this in ten minutes, ten months and ten years? Three horizons separate the anxiety from the actual stakes, and it finishes by asking what you'd tell a friend in exactly your position. Most decisions shrink.

05

Research & learning

Getting to competent on an unfamiliar subject in an evening — without quietly absorbing confident nonsense along the way.

Prompt 21

Any subject, one page

The one-sentence version, the five concepts everything is built on, the mental model that makes it click, the three things beginners get wrong, and what to learn next. One concrete example runs all the way through so it doesn't collapse into five disconnected definitions.

Prompt 22

The primer before the deep dive

The twenty terms you'll hear in an unfamiliar room, grouped by theme with the five that actually matter starred — plus three questions that make you sound like you know what you're doing. Vocabulary first is the fastest route into any new field.

Prompt 23

Compare like an analyst

You describe your use case properly and it picks the six criteria that matter for you, then scores each option with reasoning, names who each option is wrong for, and surfaces the hidden costs. Letting it choose the criteria is the point — you find out what you should have been comparing on.

Prompt 24

Steelman both sides

The strongest honest case for and against a claim, with the weakest link in each, what each side would have to concede, and where they actually agree. It's explicitly told not to retreat into both-sides mush when the evidence genuinely favours one.

Prompt 25

Interrogate a document

Turns AI into an analyst for a contract, report or brief: what matters, what would change a decision, stated versus implied numbers, and what's conspicuously missing — citing the section each time, and writing "not stated" instead of filling gaps. That last rule is your hallucination guard.

06

Planning & getting unstuck

Turning the things you've been avoiding into things that have an obvious first step.

Prompt 26

Vague goal to this week's plan

Works backwards from a deadline to milestones, then gives you only this week's tasks — maximum five, each under ninety minutes, in order. The first one gets broken into its first three physical actions, which removes the last excuse.

Prompt 27

The ten-minute start

For anything you're procrastinating on: the smallest possible first action that leaves visible progress, then the next one. It's explicitly told not to motivate you, which saves three paragraphs of pep talk you were going to skip.

Prompt 28

Unstick me (questions first)

Before giving any advice, it has to ask you the five questions that would most change its recommendation, and wait. This is the single highest-leverage upgrade in the pack — it turns a generic answer into a tailored one, and you can bolt it onto almost any request.

Prompt 29

A checklist someone else could run

Turns a project into steps a competent stranger could execute without asking you anything: verb first, one action each, with the "done" condition stated and [DECISION NEEDED] flags where you're still required. Written this way it doubles as your delegation brief.

Prompt 30

The weekly review

What genuinely moved the needle, what was busywork dressed up as work, the one thing you avoided all week, three ranked priorities, and one thing to stop doing. It's told not to congratulate you. The avoided item is usually where the useful answer lives.

07

Sales, clients & money

The prompts with a number attached to them — proposals, pricing, objections and the conversations that protect your margin.

Prompt 31

The proposal that answers the real question

A one-page proposal that leads with the client's outcome rather than your process, with numbered deliverables, a timeline, the price and what it's based on — and an explicit "not included" section, which is what prevents the scope argument in week three.

Prompt 32

Handle the objection before it's said

Eight objections ranked by likelihood, including the two nobody says out loud — price anxiety and fear of looking foolish internally. For each, the one-line answer and the proof needed for it to be believed. Then it rewrites your offer so the top three never come up.

Prompt 33

Price the thing

Asks for three pricing structures rather than three numbers, so you get retainer versus project versus outcome-based, the psychology of each, and what each signals about you — plus the number to test first and how you'd know within a month that it's wrong.

Prompt 34

The scope-creep reply

Says yes to the outcome, names the request clearly as additional scope, gives the cost and time, and offers a choice: add it, swap it for something already in scope, or do it later. Friendly, zero resentment, no free work.

Prompt 35

The update they don't have to chase

Shipped, in progress with blockers, needs-you items with deadlines, next week — under 150 words and leading with anything that's late rather than burying it. Leading with the bad news is exactly why this one builds trust instead of spending it.

08

Data & spreadsheets

Where a five-minute prompt replaces forty minutes of forum searching and one very confident wrong answer.

Prompt 36

The formula, explained

You describe what should happen and where your data lives, and get the formula, one plain-English line on what it does, the two ways it will break with guards, and the Excel variant. The "how it breaks" part is what saves you next month when someone inserts a blank row.

Prompt 37

Audit this messy data

Paste ten sample rows and get every problem named: inconsistent formats, duplicates, blanks, mixed types, dates stored as text — each with the fix and whether it's safe to automate. It's told not to fix anything yet, because bulk-editing data you haven't inspected is how columns get destroyed.

Prompt 38

What is this data actually telling me?

Five observations ranked by how much each would change your decision, the one metric you're not tracking, and anything that looks like a data problem rather than a real finding. Ranking by decision-impact stops it reporting that your biggest month was your biggest month.

Prompt 39

Build me the tracker

Designs a sheet end to end: tabs, exact column headers, which columns are inputs versus formulas (with the formulas), three summary metrics and the conditional formatting rules — optimised explicitly for you actually updating it every week rather than abandoning it in March.

Prompt 40

Sanity-check my numbers

Checks the arithmetic, then the assumptions: which one the result is most sensitive to, which is most likely wrong, and what a 30% miss does to the bottom line. Every model has one load-bearing assumption and this finds it.

09

Content & social

One idea worked properly beats five ideas half-posted. Five prompts for getting more out of what you've already made.

Prompt 41

One idea, five formats

Turns a single idea into a LinkedIn post, a thread, a sixty-second script, an email and a one-line version — each written native to its format rather than the same paragraph resized five times. That instruction is the difference between repurposing and copy-pasting.

Prompt 42

Hooks that aren't clickbait

Fifteen opening lines with the clichés explicitly banned — no "in today's world", no rhetorical questions — and a rule that every hook must contain a number, a named pain, a contradiction or a concrete image. Banning specific phrases works far better than asking for something strong.

Prompt 43

Long thing into short things

Pulls the eight most quotable standalone ideas out of a transcript or article — things that make sense with zero context — with a tightened version of each and the format it suits. One long piece is a month of short ones, which is the part most people skip.

Prompt 44

The comment reply pack

Groups comments by what the person actually wants (agreement, question, objection, self-promo, bad faith), drafts replies to the ones worth answering in your voice, tells you what to ignore, and flags which comment deserves its own post.

Prompt 45

The content audit

What your top five posts have in common that the bottom five don't — format, hook type, topic, length, timing — what you keep doing that doesn't work, and what to test next. It's told to ignore sample-size noise, so you don't get confident conclusions drawn from three data points.

10

Life admin & everyday

The non-work half of the day, where most people never think to use any of this at all.

Prompt 46

A meal plan that survives the week

Seven dinners with cook times, a shopping list grouped by supermarket section, Sunday prep, and two meals that deliberately reuse leftovers. The rule that no ingredient may appear in only one meal is what stops half a bunch of herbs rotting in the fridge every week.

Prompt 47

Explain this document to me

Plain-English breakdown of a contract, policy or official letter: what you get, what you must do, every deadline, anything unusual or one-sided, and three questions to ask before signing — with the exact clause quoted for each point so you can check the reading yourself.

Prompt 48

A trip that isn't a list of landmarks

Day-by-day and grouped by neighbourhood so you're not crossing a city twice, with one anchor per day, nearby options, where to eat, a wet-weather backup, and what must be booked ahead. Geography over ratings is the difference between a holiday and a commute.

Prompt 49

The hard conversation

Your opening two sentences, the one thing you must say clearly however it goes, three things they might say with non-escalating responses, and what to do if it goes badly — plus what not to say even though you'll want to. That last one is what people report back on.

Prompt 50

Buy the right one

The four specs that actually matter for your use case and which ones are marketing, three options at different prices, what you'd regret about each, and the cheapest option that would still make you happy. Reviews almost never lead with the regret.

The download

Take all 50 with you.

The page above tells you what each prompt does. The PDF gives you the actual templates — plus everything that didn't fit here.

  • All 50 prompts as full copy-paste templates, blanks marked
  • A short "why it works" note under every single one
  • The upgrade table: 10 weak lines and what to write instead
  • Troubleshooting: 10 ways AI output goes wrong, and the fix line for each
  • A reusable context block to paste into your AI's custom instructions
  • 10 one-liners to bolt onto the end of any prompt you write yourself
Free · 31-page PDF

Send me the 50 prompts

One email field. The download appears immediately after — no confirmation loop, no waiting.

No spam. Unsubscribe in one click. You'll also get the weekly build.

Questions

Before you download

Is the prompt pack really free?

Yes. Enter your email, confirm, and the PDF downloads immediately — no card, no trial, no upsell page in between. You'll also get the weekly Steal This System newsletter, and you can unsubscribe from any issue in one click.

Which AI tools do these prompts work with?

All of them. Every prompt is written to be model-agnostic and works in ChatGPT, Claude, Gemini, Microsoft Copilot and anything else built on a general-purpose model. Nothing in the pack needs a paid plan, and where a prompt performs better with web search enabled, the pack says so.

Do I need any technical or prompt engineering experience?

No. Each prompt is a fill-in-the-blank template — anything in [square brackets] is yours to replace, and everything else stays as written. The pack also teaches the five-block formula behind every prompt, so by about page seven you can write your own instead of hunting for lists.

How is this different from the free prompt lists everywhere else?

Most lists give you one-line suggestions like "act as a marketing expert". These are complete working templates with the output format, constraints and quality bar already specified — the parts that actually determine whether the answer is usable. Each one also explains why it works, and there's a troubleshooting section for the ten ways AI output goes wrong.

What format is the download?

A 31-page PDF, designed for both screen and print, with selectable text so you can copy any prompt straight out of it. It works on desktop, tablet and phone.

Can I share it with my team?

Please do. Free to share, never to sell. If it's useful to a colleague still typing "write me a professional email", send it on.