The Briefing Formula: five ingredients, two power moves, no magic words

Yesterday you had your first real conversation and met the Confident Intern. Today you learn the skill the whole course stands on: the difference between a question and a briefing. By the end you'll have written one real, sendable email — and you'll know why it worked.

1Whatever you don't say, it guesses

Here is the single most useful fact about every AI assistant you will ever use: it knows nothing about your situation except what's in the conversation. It has read half the internet, yes — but it has not read your inbox. It doesn't know your landlord's name, how long your heating has been broken, whether you've already called twice, or whether you want to stay friendly or escalate. The chat window is its only window into your life. (Some assistants now keep a light "memory" of past chats — you may see ChatGPT recall an earlier preference. Treat that as a bonus, never a guarantee: for facts that matter, the only context you can count on is what you put in the conversation.)

So when your prompt leaves a gap, the assistant doesn't stop and raise its hand. By default, it fills the gap — with the most typical, most average choice it has seen a million times. Average length. Average tone. Average assumptions. That's what "generic" actually is: not a malfunction, but an average. The answer isn't wrong, exactly. It's nobody's. And sometimes it's worse than nobody's, because to sound complete, the assistant will invent the details you didn't supply — the Confident Intern from Day 1, confidently filling in a history that never happened.

Flip that around and you get today's whole lesson: the quality of the answer is decided before you press send. When people say "AI gives me generic answers," what's nearly always true is that they gave it a generic briefing. Hand a brilliant intern a sticky note that says "email landlord heating" and you'll get something grammatical and useless. Sit them down for sixty seconds and explain the situation, and you'll get work you can actually use. Same intern. Same brain. Different briefing.

And the trade is absurdly in your favor. A complete briefing costs you roughly one extra minute of typing. An incomplete one costs you three rounds of "no, not like that," a result you quietly rewrite yourself, and the creeping suspicion that AI is overrated. One minute, against all of that. Today you learn what goes into that minute.

2The Briefing Formula: Role · Goal · Context · Format · Tone

A briefing is a prompt with five ingredients. You already know all five — you use them instinctively whenever you hand work to a human being. The formula is only a checklist that stops you from leaving one out:

• Role — who the assistant should be for this task. It sets the perspective and the vocabulary.
• Goal — the one thing you want back, phrased as an action: write, compare, plan, rewrite, list.
• Context — the facts of your situation. The ingredient beginners skip most, and the only one the assistant cannot guess.
• Format — the shape of the answer: length, structure, bullets or prose, a subject line, a table.
• Tone — how it should sound: warm, firm, playful, formal, plain.

Watch what the formula does to one ordinary request. On the left, the prompt most people send. On the right, the same request as a briefing:

Now each ingredient on its own, with what it buys you — and, more importantly, what leaving it out costs you. This is the Briefing Stack; it's worth a slow read, because every prompt for the rest of the course is built from these five blocks:

Two practical notes before we use it. First: you don't need all five every single time. For a tiny task, Goal plus Context often carries it. But the moment the output matters — anything a real person will read — run the full stack; the missing ingredient is always the one that bites. Second: the order doesn't matter, completeness does. Write in full sentences, give real context, and say what shape and sound you want. That's the course-wide style from here to Day 10.

3The two power moves

The formula covers what you know to say. The first power move covers what you don't know you forgot — and it is the single highest-leverage line a beginner can learn. End your briefing with:

Before you answer, ask me up to 3 questions that would make this better.

This is "ask me first," and it quietly reverses the roles: the assistant becomes the interviewer. You can't list context you don't realize is missing — but the assistant, having seen a million versions of your task, knows exactly where the gaps usually are. Its questions are a map of your missing ingredients. Is there a deadline? What was agreed on the phone? How firm do you want to be? You answer in plain sentences, and the briefing repairs itself. It's the same thing every good professional does: a serious freelancer asks questions before quoting; a serious doctor asks questions before prescribing. Now your assistant does too — because you asked it to.

The second power move: iterate — the first answer is a first draft. Even after a perfect briefing, treat what comes back the way you'd treat an intern's first attempt: useful, promising, and not finished. Don't start over. Don't open a new chat. Reply, in the same thread, with feedback:

• "Make it half as long."
• "Warmer. Less formal."
• "Give me 3 versions: friendly, neutral, firm."
• "Keep the first paragraph, redo the rest."
• "You've lost the part about the phone call — put it back."

Remember from Day 1: a thread remembers what was said inside it. Every fact you've established, every preference you've expressed, is still on the table — so each round of feedback compounds instead of starting from zero. Iteration isn't a sign the tool failed. Iteration is the workflow. The people who get remarkable results from these assistants aren't writing remarkable first prompts; they're having short, demanding conversations.

4No magic words — and your formula travels

One de-mystification, because the internet will try to sell you the opposite. You will meet lists titled "300 ChatGPT prompts that will change your life." You don't need them. There is no secret vocabulary, no hidden incantation, no phrase that unlocks a better model. These assistants are built to follow clear instructions written in ordinary language. What looks like wizardry from the outside is, every single time, completeness: a role, a goal, real context, a format, a tone. There are complete briefings and incomplete ones — that is the entire taxonomy of prompting.

And here's the part that makes today's hour a genuinely durable investment: prompts are portable across assistants. The Briefing Formula isn't ChatGPT knowledge or Claude knowledge — it's communication. The same briefing works in ChatGPT, Claude and Gemini, because all three are built to do the same fundamental thing: turn a clear request into useful work. Tomorrow you'll compare the three and pick your daily driver; whichever you choose, today's skill comes with you untouched. Interfaces get redesigned, models get renamed, tools come and go — the briefing survives all of it, because it lives in your head, not in their menus.

1. Open ChatGPT and start fresh. Go to chatgpt.com and click the new-chat button (usually at the top of the sidebar, next to your chat history). an empty message box at the bottom of the screen, waiting.
2. Feel the baseline. Send the flat prompt exactly as it is — no improving it yet. We need the "before" picture.
   Copy-paste prompt · the flat versionWrite an email to my landlord about the broken heating.
   a competent but hollow email. Look closely — it either invented specifics (how long it's been broken, what you've done about it), littered the text with [placeholders] for you to fill in, or asked you a question before writing. All three are the same signal: it had to guess, because you didn't brief.
3. Send the same request as a full briefing. The labels in parentheses are training wheels for you — the assistant doesn't need them, and you can drop them once the five ingredients are reflex. If you have a real situation, swap in your own context line.
   Copy-paste prompt · the full briefingYou are an experienced assistant who knows tenant communication well. (ROLE) Write an email to my landlord asking for the heating to be repaired this week. (GOAL) Context: it has been broken for 5 days, I reported it by phone on Monday, I have a 2-year-old at home, and I want to stay on good terms. (CONTEXT) Format: under 150 words, with a subject line. (FORMAT) Tone: polite but firm, no legal threats. (TONE)
   a specific, sendable email with your facts in it — the Monday call mentioned, the 2-year-old doing its quiet work, no invented history, the right length and temperature. Usually on the first try.
4. Power move 1 — make it interview you. New message, same chat — this time on a task you've barely briefed, so it has real gaps to find. Send:
   Copy-paste prompt · ask me firstI also need to message my neighbor about the noise last weekend. Before you write anything, ask me up to 3 questions that would make this message better.
   Answer its questions in plain, ordinary sentences — one or two lines each is plenty. two or three sharp questions — how bad was it, have you spoken before, friendly note or formal complaint? Each one is a missing ingredient it found for you. This is the move that separates briefing from guessing.
5. Power move 2 — iterate in the thread. Still in the same conversation:
   Copy-paste prompt · three versionsGive me 3 versions: friendly, neutral, firm.
   Read all three, then push once more:
   Copy-paste prompt · tighten the winnerTake the neutral one and make it 30% shorter.
   three distinct takes, then a tightened neutral version — and notice what you did NOT have to do: re-explain anything. The thread carried every fact from steps 3 and 4. That's iteration compounding.
6. Prove the portability claim. Open claude.ai in another tab (you made the account on Day 1), start a new chat, and paste the exact briefing from step 3 — unchanged. a different voice — perhaps a touch more formal, perhaps warmer — but the same usefulness, your same facts, a same-quality email. The formula was never the tool's. It's yours, and it travels.