Ask So It Lands — Party School

Core concept 01

A good ask is a handoff, not a search.

When you type into a free AI tool like Claude or ChatGPT, the message you send has a name: it is called a prompt. That is just the industry word for your ask, nothing fancy. And the single biggest shift you can make is to stop treating the prompt like a Google search and start treating it like handing a task to a capable new assistant. The tool knows an enormous amount about the world. It knows nothing about your world yet, your clients, your voice, what you actually needed made, until you say so.

So "write me a post" lands the way it would land on a real person's first morning: they would smile, guess, and hand you something technically correct and completely generic. Anthropic, one of the companies that builds these models, puts the fix bluntly in its own guidance: tell the model exactly what you want to see. A weak answer is almost never the model being bad at its job. It is a briefing that was too thin to aim at.

Source: Anthropic, "Prompt engineering best practices for 2026"

Core concept 02

The five-part ask: Who, What, Context, Example, Format.

Both of the big labs, Anthropic and OpenAI, describe the same handful of ingredients for a prompt that works. We are going to call the gentle version the five-part ask, and you can hold the whole thing in your head:

Who the helper is for this task. What you want made, named plainly. The Context from your world it needs to know. One real Example of good. And the Format, the shape you want back. Each part exists because it answers a question the model would otherwise have to guess at, and a guess is exactly how you end up with generic.

Of the five, one does the most work: the Example. Anthropic says it plainly, examples show rather than tell, and they carry the hard-to-describe stuff like tone and structure better than any amount of explaining. One thing you were proud of, pasted in with "make it sound like this," teaches the model your taste in a way three paragraphs of description never will.

Sources: Anthropic, "Prompting best practices" · OpenAI, "Prompt engineering"

The five-part ask · Who, What, Context, Example, Format

Core concept 03

Watch a weak ask become a strong one.

Here is the whole lesson in one before-and-after. The weak ask is one you have probably typed. The strong ask is the exact same request with the five parts added, and it is not longer because you are trying harder, it is longer because you stopped making the model guess.

Weak: "Write me a post."

Strong: "You are helping me run my photography studio (Who). Write an Instagram post announcing fall mini-sessions (What). I have eight slots, they are 30 minutes each, first weekend of October (Context). Here is a past post that got a great response, match this warmth: [paste] (Example). Keep it under 80 words, friendly, one clear next step (Format)."

Every part earns its keep. Who puts the model in your shoes instead of a generic one. What names the actual thing so it stops writing "about" your topic and starts making the thing. Context hands over the real details only you know. Example carries your voice. Format saves you the round of "shorter, warmer, drop the emoji." Anthropic's guidance adds one gentle bonus: saying why it matters, "these slots always sell out and I want it to feel like a heads-up to friends," helps the model aim at your real goal, not a textbook one.

A little more if you're curious: you do not need all five, every time

The five parts are a checklist, not a rule you have to fill out like a form. For a quick task, Who and What alone are a big upgrade over "write me a post." The moment an answer comes back a little off, that is your cue to add the part that was missing, usually Context or Example. You will get a feel for which one fast, and the next section is a cheat sheet for exactly that.

And you never have to nail it in one shot. Real pros go back and forth: they read the first draft, say "warmer, and cut the last line," and let the model try again. Correcting out loud in plain words is not failing at prompting. It is prompting.

Source: Anthropic, "Prompting best practices"

Scroll the diagram sideways →

Before and after · a thin answer versus one you can actually use

Core concept 04

Read the weak answer. It tells you the part you skipped.

A flat answer is not a dead end, it is a clue. The way it misses points straight at the part of the ask you left out, and the fix is almost always one line you add, never a rewrite from scratch. Here is the whole cheat sheet.

Read your answer · the ask, the missing part, the fix
You asked	The missing part	The line to add
"Write me a post." The answer could be anyone's.	Who + What	Say who it's for and what it's about: "a post for my studio's followers announcing fall mini-sessions."
"Write a gallery email." It's true, but empty of your details.	Context	Hand over your world: "12 fall weddings booked, galleries go out about 3 weeks after the day."
"Make it sound like me." Close on words, wrong on voice.	Example	Show one: paste a past email you loved and say "match this." Voice is caught, not described.
"Write a warm client note." Right idea, wrong length or shape.	Format	Name the shape: "3 sentences, one link, no exclamation points."
"Answer this customer." It explained instead of drafting a reply.	What	Name the deliverable: "draft a reply that offers a reschedule, ready for me to send."

Notice that not one fix is "try harder." Every single one is a small thing you add. That is the whole reason a better ask is learnable in an afternoon: it is almost entirely additive, and you already know all the details, you just had not said them yet.

The ask you leave with

Fill in the five-part ask for one real task

Pick one thing you write over and over, the email, the caption, the reply you retype every week. Fill in these five blanks for that exact task. Paste the finished thing into a free Claude or ChatGPT chat and watch what comes back. This is your first real ask, built from your own work.

01Who am I"You are helping a [your role] with [the corner of your work this task lives in]."
02What I wantthe thing, named plainly: "write the [email / caption / reply] that [does what]."
03Context I'll pastethe real details only you know: names, numbers, dates, the situation this time
04One examplea past piece you were proud of, pasted in, with "make it sound like this," the one line that does the most work in the whole ask
05The formatlength, tone, and any hard rules: "[N] words, [tone], [one link / no emoji / your rule]"

Fresh from the lab

What changed this month

Anthropic told its own team to delete instructions, not add more. Anthropic makes Claude. On July 24 the team behind Claude's coding tool said they cut more than 80% of the standing instructions built into it, and the tool did not get worse, it stayed just as good. Those instructions were extra rules written for an older, less capable version of the model. The newer model already knows better, so the old rules were just clutter sitting in its way. Anthropic even built a checkup tool that looks at your own instructions and tells you what is safe to cut.

OpenAI's new guide for its newest model says the same thing in different words: say what you want, not how to do it. Published July 9 for ChatGPT's GPT-5.6 family, the guidance says to state your goal and your limits, then let the model figure out the steps itself instead of walking it through each one. In OpenAI's own tests, shorter, goal-first prompts scored 10 to 15% better and needed up to two-thirds fewer words to say it, which means faster answers and a smaller bill.

Notice the pattern in both: less can genuinely be more once you trust the model with the goal.

Sources: Anthropic, "The new rules of context engineering for Claude 5 generation models", July 24, 2026 · TechTimes, "GPT-5.6 Prompting Guide", July 15, 2026

Plain vocabulary

Six words, each with a plain translation

You do not need to memorize these to use the five-part ask. But you will see them in AI menus and articles, so here is each one in plain words, no prior knowledge assumed.

Prompt

The message you type to an AI. That is the whole definition. Your ask, your question, your instructions, all of it is "the prompt." When someone says "write a better prompt," they mean "ask more clearly."

Context

The background you hand over so the answer fits your world: your numbers, your situation, the files you paste in. More of the right context is the difference between a generic answer and yours.

Example

A sample of what "good" looks like, pasted right into the ask. Insiders call this a "few-shot" example, a fancy phrase for "here are a few I liked, do it like these." It is the strongest move you have.

Format

The shape you want the answer in: how long, what tone, a list or a paragraph, any hard rules like "no emoji." Saying the format up front saves you a round of "shorter, please."

System prompt

A standing set of instructions the AI reads before every message, so you do not retype who you are each time. In free Claude and ChatGPT this is the "instructions" or "custom instructions" box you fill in once.

Model

The actual AI doing the work, with a name and a version, like Claude or GPT-5.6. The free tools run a capable model at no cost. You do not need the paid one to do everything in this lesson.

Deeper definitions, straight from the builders: Anthropic prompt engineering docs · OpenAI prompt engineering guide

Make it yours

Write one genuinely good ask

No kit to build, no account to upgrade. Pick your lane and write one five-part ask for the task you do most, then send it in a free Claude or ChatGPT chat. One good ask, start to finish, is the whole win today.

For female founders

I'm the CEO, the marketer, the bookkeeper, and the intern, and I write every proposal, post, and follow-up from a blank page because I've never saved a version of my own voice.

One good ask: this week's follow-up email

Tool: a free Claude or ChatGPT chat · no paid plan, no setup

Who: start your message with "You are helping me, the founder of [company], write a follow-up email." One sentence, and the model is already in your shoes.
What: name the exact thing, "write the follow-up to the customer I talked to yesterday," not "help me with emails."
Context: paste the real bits, who they are, what you discussed, the one ask you want to make. This is the part only you can give.
Example: paste one past email you were proud of and add "match this tone and length." This one line does the most work.
Format: end with "under 150 words, warm, one clear ask, no jargon." Send it, read the draft, and say what to fix in plain words.

The payoff: one real email, done in minutes, that actually sounds like you and not a template.

For artists

I make the work, then the statements and applications eat the studio time, and every single one starts from zero in my own voice.

One good ask: the caption for your newest piece

Tool: a free Claude or ChatGPT chat · no paid plan, no setup

Who: open with "You are helping me, a [medium] artist, write about my own work in my voice."
What: "Write a short social caption for the piece I just finished." Name it, don't just say "help with captions."
Context: tell it about the piece in a sentence or two, what it is, what it's about, anything true you'd want a viewer to know.
Example: paste your strongest past caption and say "this length, this directness, and never use 'explore' or 'juxtapose' unless I do first."
Format: "two or three sentences, plain and specific." Read it, cut anything that sounds like a brochure, ask again.

The payoff: a caption that sounds like you, written in a coffee break instead of a lost studio hour.

For actors

Between survival jobs, self-tapes, and submissions, I retype the same cover note a dozen slightly worse ways because I never saved the one that actually booked something.

One good ask: today's submission cover note

Tool: a free Claude or ChatGPT chat · no paid plan, no setup

Who: start with "You are helping me, an actor, write a submission cover note in my voice."
What: "Write the cover note for this specific breakdown," and paste the breakdown right in.
Context: add the one credit or detail that fits this role best. Give it the thing a casting director should notice first.
Example: paste the cover note that actually got a callback and label it "the standard, warm and specific, never desperate."
Format: "short, three or four sentences, lead with the credit that fits." Read it, adjust, resend.

The payoff: a real submission out the door in minutes, still in your voice, not a generic one.

For musicians

I'm a musician, not a manager, but I'm the one writing gig pitches, deposit chasers, and midnight replies to "do you do weddings?" with no template in sight.

One good ask: the reply to a booking inquiry

Tool: a free Claude or ChatGPT chat · no paid plan, no setup

Who: "You are helping me, a working musician, reply to a booking inquiry in my voice."
What: "Write the reply to this message," and paste the actual inquiry that just came in.
Context: your real details, what you play, your rate range, your one clear next step, so the reply is true and not made up.
Example: paste the reply that actually booked a room and say "talk like I talk from stage, this exact tone."
Format: "three sentences, one clear ask, one link." Read it, cut anything too polished, resend.

The payoff: the midnight inquiry answered before you forget it, and it still sounds like you.

For fitness pros

I'm teaching back to back all day, and by 9pm I'm still writing captions, DMs, and re-sign nudges from scratch every time.

One good ask: tonight's client check-in reply

Tool: a free Claude or ChatGPT chat · no paid plan, no setup

Who: "You are helping me, a fitness coach, reply to a client's check-in in my voice."
What: "Write my reply to this client's update," and paste their actual note.
Context: add what you'd want to say back, the one cue or adjustment, and a hard rule: "never invent an exercise I don't coach."
Example: paste your best real check-in reply and say "match this warmth."
Format: "three sentences max, warm and direct." Read it, fix anything off, resend.

The payoff: the 9pm reply done in two minutes, warm as ever, and your evening back.

For coaches

I sell transformation, but I spend my week rewriting the same follow-up email and recap note because I never turned the good one into a template.

One good ask: today's session recap

Tool: a free Claude or ChatGPT chat · no paid plan, no setup

Who: "You are helping me, a coach, write a session recap in my voice."
What: "Write the recap email for today's session," named as the exact deliverable.
Context: your notes from the session, with a firm rule you type every time: "no client's real name, ever."
Example: paste a recap a client actually thanked you for and say "this level of specific."
Format: "the recap, one action step, one question." Read it, tune it, send it same day.

The payoff: the recap goes out the same day, instead of onto the someday pile.

For therapists

Between sessions I'm drafting intake replies and insurance letters from a blank page, and none of it is why I trained.

One good ask: an intake FAQ reply, with a bright red line

Tool: a free Claude or ChatGPT chat · admin only, never clinical

Draw the line first: no client information goes into a general AI tool, ever. This ask is about your practice, never a person.
Who: "You are helping me, a therapist, write a warm front-office reply to a new inquiry."
What + Context: "Write a reply that answers a prospective client's common questions about fees, availability, and next steps," and give the real, general facts.
Example: paste a past intake reply you liked, with any identifying detail removed first, and say "match this."
Format: "warm, clear, a few short paragraphs, never reference a specific person."

The payoff: the between-session admin shrinks, and the tool never touches anything clinical.

For authors

I want to be writing the next book, and instead I'm rewriting the same launch email and pitch note from scratch every single time.

One good ask: this month's newsletter

Tool: a free Claude or ChatGPT chat · no paid plan, no setup

Who: "You are helping me, an author, draft my reader newsletter in my voice."
What: "Write this month's newsletter," named, not "help me with my list."
Context: your one update and the one link you want in, plus a note: "how I talk to readers in the book's acknowledgments."
Example: paste the past issue with the best replies and say "this structure."
Format: "about 300 words, one story, one link." Read it, cut what isn't you, resend.

The payoff: the newsletter drafted in one sitting, and mornings back on the book.

For chefs

I'm cooking or shopping all day, so every quote and menu description gets written cold, with no starting point to work from.

One good ask: tonight's special, described

Tool: a free Claude or ChatGPT chat · no paid plan, no setup

Who: "You are helping me, a chef, write a menu description in my voice."
What: "Write the menu line for tonight's special," named plainly.
Context: the dish, the key ingredient, where it's from, the price, plus a rule: "no adjectives I wouldn't say out loud across the pass."
Example: paste the menu line guests actually asked about and say "this exact style."
Format: "dish name, a three-word story, price." Read it, trim it, print it.

The payoff: the special written in the time it takes to plate one, and it sounds like your kitchen.

For consultants

I'm delivering client work all day, so my proposals and recap emails start from a blank page every time, right when I have the least time to write well.

One good ask: today's call-recap follow-up

Tool: a free Claude or ChatGPT chat · no paid plan, no setup

Who: "You are helping me, a consultant, write a call-recap follow-up in my voice."
What: "Write the follow-up to today's call," named as the deliverable, ready to send.
Context: paste your notes, the problem you heard, and the next step you want to propose.
Example: paste a follow-up that actually moved a deal and say "this structure, credible and unhurried, no jargon."
Format: "problem, approach, one clear next step." Read it, adjust, send.

The payoff: the follow-up out the same afternoon, which is the whole trick of keeping a pipeline warm.

For health advocates

Every client is a crisis and the paperwork is endless, and I'm drafting intake notes and appeal letters from zero each time.

One good ask: an appeal-letter skeleton, privacy first

Tool: a free Claude or ChatGPT chat · templates only, never case details

Privacy rule first: no client health information in a general AI tool, ever. Build a reusable skeleton with [BRACKETS], and fill the person's details in yourself, after.
Who: "You are helping me, a patient advocate, write a reusable insurance-appeal skeleton."
What + Context: "Write an appeal template with a denial-reason section, a counter-argument section, and the ask, using [BRACKETS] for anything specific."
Example: paste a de-identified appeal that actually won and say "this firmness, plain, no legal jargon."
Format: "denial reason, counter-argument, the ask, all bracketed." Save the wording you like for next time.

The payoff: the 11pm letter becomes a fifteen-minute fill-in-the-blanks job, with no private detail ever in the tool.

For makers

I make the thing, then I write the listing, the caption, and the wholesale pitch from scratch for every single piece.

One good ask: the listing for the piece on your bench

Tool: a free Claude or ChatGPT chat · no paid plan, no setup

Who: "You are helping me, a maker, write a product listing in my voice."
What: "Write the listing for the piece I just finished," named, not "help with my shop."
Context: the materials, the process, the price, and one true detail about this piece.
Example: paste your best-selling listing and say "this exact structure, specific and unfussy, no adjective that could describe any product."
Format: "title, two-sentence story, materials, care." Read it, tidy it, post it.

The payoff: a real listing done in one sitting, and the studio gets its maker back.

For nonprofit leaders

I'm chasing grants and thanking donors, and every letter starts from a blank page even though I say the same true things every time.

One good ask: this week's donor thank-you

Tool: a free Claude or ChatGPT chat · no paid plan, no setup

Who: "You are helping me, a nonprofit director, write a donor thank-you in my voice."
What: "Write the thank-you to a donor who just gave," named as the deliverable.
Context: the gift, one real thing it makes possible, and one true impact number.
Example: paste language from a letter that landed well and say "concrete, warm, never grandiose."
Format: "short, one real number, one specific thanks." Read it, adjust, send.

The payoff: the thank-you out within the day, so development stops eating the mission's hours.

For photographers

I'm booked shooting on weekends and editing all week, so every inquiry reply and gallery email gets written cold, mid-scramble.

One good ask: today's inquiry reply

Tool: a free Claude or ChatGPT chat · no paid plan, no setup

Who: "You are helping me run my photography studio's front desk, in my voice."
What: "Write the reply to this inquiry," and paste the actual message that came in.
Context: your real availability, packages, and the next step, so nothing is made up.
Example: paste the inquiry reply that actually booked and say "this exact warmth."
Format: "warm, in my voice, real next steps only." Read it, adjust, send same day.

The payoff: inquiries answered the same day, without a VA and without the scramble.

For realtors

My database is a graveyard, and every listing description and nurture email starts from zero instead of from what already worked.

One good ask: this week's listing description

Tool: a free Claude or ChatGPT chat · no paid plan, no setup

Who: "You are helping me, a realtor, write a listing description in my voice."
What: "Write the description for this listing," named plainly.
Context: the three features that actually sell it and the neighborhood detail buyers care about.
Example: paste the listing that sold fastest and say "sound like a neighbor, not a brochure, this exact tone."
Format: "150 words, three features, one call to come see it." Read it, tune it, post it.

The payoff: a ready-to-post description in minutes, built from what already worked, not a blank box.

For stylists

I'm behind the chair all day, so every rebooking text and caption gets typed out fresh, one client at a time, from nothing.

One good ask: today's rebooking text

Tool: a free Claude or ChatGPT chat · no paid plan, no setup

Who: "You are helping me, a hairstylist, text a client between appointments, in my voice."
What: "Write a rebooking text for a client whose usual window is coming up."
Context: the one specific detail that makes it personal, their service, their timing, so it doesn't read like a blast.
Example: paste a rebooking text that actually worked and say "sounds like a text, not an ad, this exact length."
Format: "one line, one specific detail, one link." Read it, adjust, send.

The payoff: the day off goes back to being a day off, and rebookings still get sent.

The house rule

What belongs to you and keeps your humanity, you keep. What you'd write roughly the same way every time, you hand to a clear ask that already knows the brief. The five-part ask is not a shortcut around your voice. It is the fastest way to get your voice out of your head and onto the page, with the boring part carried for you.

Next lesson: Pick your tools →