The 5-Line AI Brief: One Template for Emails, Ads, and Negotiations
Published 2026-06-09
Most disappointing AI answers have the same root cause: the request was vague, so the model had to guess. It guessed the tone, guessed the audience, guessed the length — and got at least one of them wrong. The fix isn't a secret incantation or a 400-word mega-prompt. It's a habit: before you hit send, spend thirty seconds answering five questions. Who should the AI be? What's the situation? What's the one outcome you want? What are the limits? What shape should the answer take?
That's the whole template. Five lines. It works for emails, ad copy, negotiation messages, meal plans, and most everyday tasks where you want a usable result on the first or second try. In this article we'll walk through each line, show why skipping it backfires, and then fill the template in for three real situations: asking for a raise, writing a local promo post, and negotiating rent.
Here's the empty template. Copy it somewhere you can reach it quickly — a note, a text expander, the back of your hand:
Role: [who the AI should be — a profession or perspective]
Context: [the situation, plus who will read the result]
Goal: [the ONE outcome you want from this output]
Constraints: [tone, length, things to avoid]
Format: [the exact shape of the output]
If you've read our piece on prompt engineering vs. briefing, you'll recognize the philosophy: you're not hacking the model, you're briefing it the way you'd brief a capable colleague who knows nothing about your situation. The five lines are just the minimum information any colleague would need.
Let's take them one at a time.
Line 1: Role — who the AI should be
The role line tells the AI which hat to wear. "You are an experienced HR manager," "You are a small-business copywriter," "You are a calm, practical tenant advisor." One short sentence is enough.
What it does: it sets the vocabulary, the priorities, and the level of caution. A career coach and a lawyer will give you very different advice about the same email, and both will differ from the generic middle-of-the-road answer you get when you specify neither.
The failure mode when it's missing: you get the average of everyone. The model defaults to a neutral, slightly corporate assistant voice that belongs to no profession in particular. The advice isn't wrong, exactly — it's just bland, hedged, and shaped for nobody.
How to fill it: name the professional you'd actually want to ask in real life, and add one trait that matters ("direct," "warm," "skeptical").
A small caveat: the role is a lens, not a costume party. "You are a wise dragon CEO" is fun but rarely improves the output. "You are a copywriter who specializes in local service businesses" does.
Line 2: Context — the situation and the audience
Context is where you hand over the facts the AI cannot guess: who you are, what's going on, and — critically — who will read the output. This is usually the longest line, and it's the one people skip most often because the situation feels obvious to them.
What it does: context turns a generic answer into your answer. "Write a raise email" produces a template. "Write a raise email — I'm a support team lead, two years in role, took over onboarding for new hires this year, my manager is data-driven and hates fluff" produces something you can actually send.
The failure mode when it's missing: the model fills the gaps with the most statistically common situation, which is almost never yours. You then spend three rounds of follow-up corrections supplying the context you could have given upfront. This is the core idea behind the clarity principle lesson: every detail you leave out is a decision you've delegated to a guess.
How to fill it: include the two or three facts that change the answer, and always name the reader — their relationship to you, and what they care about.
A useful test: if a stranger read only your context line, could they roughly picture your situation? If not, add one more fact.
Line 3: Goal — the one outcome
The goal line states what success looks like, in one sentence, with one outcome. Not "write something about my raise" but "get my manager to agree to a salary conversation this month."
What it does: it gives the AI a decision rule. When the model has to choose between being thorough and being persuasive, between covering five points and hammering one, the goal tells it which trade to make.
The failure mode when it's missing: the output tries to do everything at once. The raise email also apologizes for asking, also recaps the whole year, also asks about remote work policy — and the actual ask gets buried in paragraph four. Outputs without a single goal read like they were written by committee, because in a sense they were: the model averaged every plausible purpose.
How to fill it: finish the sentence "after they read this, I want them to ___" — and resist the urge to add "and also."
If you genuinely have two goals, write two briefs. Two short focused outputs beat one muddled one every time.
Line 4: Constraints — tone, length, and what to avoid
Constraints are the guardrails: how it should sound, how long it can be, and what must not appear. "Confident but not entitled. Under 150 words. Don't mention my coworker's salary. No exclamation marks."
What it does: it pre-empts the most common revision requests. Most "make it shorter," "make it less formal," "remove the cheesy opener" follow-ups are constraints you knew before you started but didn't write down.
The failure mode when it's missing: defaults take over, and AI defaults run long, enthusiastic, and slightly salesy. You get three paragraphs where one would do, an opener like "I hope this email finds you well," and a tone calibrated for a press release rather than a message to a human you know.
How to fill it: write down the first two things you'd otherwise complain about in the draft — usually that's length and tone — plus one explicit "do not."
The "what to avoid" part deserves special love. Negative constraints ("no jargon," "don't apologize," "never promise a specific discount") are some of the highest-value words in any brief, because the model can't infer them from anything else you wrote.
Line 5: Format — the exact output shape
Format describes what the deliverable physically looks like: "an email with a subject line," "a 3-slide outline with one headline per slide," "a table with columns X, Y, Z," "the message text only, no commentary."
What it does: it makes the output copy-paste ready. The difference between "some thoughts on your promo post" and "the finished promo post, 2 short paragraphs plus a closing line with the offer" is the difference between homework and a deliverable.
The failure mode when it's missing: you get an essay when you needed a list, a list when you needed a paragraph, or — the classic — a helpful preamble ("Great question! Here's a draft you could consider...") wrapped around the thing you actually wanted. Structure is the cheapest quality win available, and it's covered in depth in the output formatting lesson.
How to fill it: describe the output as if you were ordering it from a print shop — components, order, and roughly how big.
The template in action
Theory is nice; filled-in briefs are nicer. Here are three complete examples you can adapt. Notice how each one stays five lines — no special tricks, just specific answers to the five questions.
Example 1: asking for a raise
Role: You are an experienced career coach who helps people negotiate
compensation without burning bridges.
Context: I'm a customer support team lead, 2 years in the role. This year
I took over new-hire onboarding and our team's response-time scores
improved noticeably. My manager, Dana, is friendly but very busy and
prefers short, concrete messages. We haven't discussed salary since I
was promoted.
Goal: Get Dana to agree to a 30-minute conversation about adjusting my
salary, within the next two weeks.
Constraints: Confident, warm, zero guilt-tripping. Under 150 words. Don't
state a specific number yet, and don't threaten to leave or hint at other
offers.
Format: A ready-to-send email with a subject line, a greeting, two short
paragraphs, and a clear ask for a meeting.
What makes this work: the goal is the meeting, not the raise itself — which is the right goal for a first email. The constraints kill the two most common raise-email mistakes (naming a number too early, sounding like an ultimatum). And because the context describes Dana, the tone gets calibrated to a real reader instead of a generic "boss."
Want to try this one yourself? It's a recurring scenario in our ask for a raise challenge — write your own brief and see how the coach scores it.
Example 2: a local ad / promo post
Role: You are a copywriter who specializes in social posts for small
local businesses — punchy, friendly, never corporate.
Context: I run a small bike repair shop in a residential neighborhood.
Spring is starting and people are pulling bikes out of storage. I'm
offering a "spring check-up" — brakes, gears, tire pressure — for a flat
$25 this month. My audience is local families and commuters on the
neighborhood Facebook group, not cycling enthusiasts.
Goal: Get neighbors to book a spring check-up this month.
Constraints: Plain, warm, neighborly tone — like a real person typing,
not a brand. Max 80 words. No hashtag walls, no "Don't miss out!!",
no cycling jargon.
Format: One Facebook post: a hook line, 2–3 short sentences, then one
line with the offer, price, and how to book.
What makes this work: the context defines the audience precisely (families and commuters, not enthusiasts), which prevents the model from writing about derailleur indexing to people who just want their kid's bike to stop squeaking. The negative constraints do heavy lifting too — without "no hashtag walls," you will get hashtag walls.
Example 3: a rent negotiation message
Role: You are a calm, practical tenant advisor who helps renters
negotiate without souring the relationship with their landlord.
Context: I've rented the same one-bedroom apartment for 3 years, always
paid on time, and handled small repairs myself. My landlord, Mr. Alvarez,
just proposed a 12% rent increase at renewal. Similar units in my
building have been listed at roughly my current rent, and I can stay
flexible on the lease length. We've always communicated politely by text.
Goal: Open a negotiation — get Mr. Alvarez to consider a smaller
increase or a longer lease at a rate closer to current, ideally over a
quick call.
Constraints: Respectful and cooperative, never legalistic or threatening.
Under 120 words, since it's a text message. Acknowledge that costs rise;
don't claim I'll move out, because I'd rather not.
Format: A single text message, ready to send, ending with a suggestion
to talk by phone this week.
What makes this work: the context includes leverage (3 years, on-time payments, comparable listings) and the relationship history, so the message can be firm without being cold. The constraint "don't claim I'll move out" is honest about the real bargaining position — and an AI that knows your real position writes a message you can actually stand behind. You can practice exactly this scenario in the negotiate rent challenge.
The same skeleton stretches to gentler situations too — a thank-you note after an interview is just a different role, a warmer goal, and a tighter length constraint.
When to break the template
The five-line brief is a default, not a law. Two cases where you should ignore it:
Quick factual questions don't need it. "What's the difference between APR and APY?" needs zero lines of role-play. Wrapping a lookup question in a five-line brief is cargo-culting — the template earns its keep only when the output has a tone, an audience, and a stake. If a wrong-ish answer costs you nothing, just ask.
Big multi-step projects need more than five lines. If you're drafting a business plan, restructuring a 30-page document, or working through a multi-day project, five lines won't hold all the necessary context — you'll want background documents, examples of work you like, and a step-by-step plan, possibly spread across several messages so you don't overload the model's context window. For those, the five lines become the opening of the brief rather than the whole thing: role, context, goal, constraints, format first, then the supporting material underneath.
There's also a middle case worth naming: iterative work. When you're refining a draft over several rounds, you don't re-send the full template each time — you send corrections ("shorter, and cut the second paragraph"). The brief sets up round one; conversation handles the rest.
A reasonable rule of thumb: if the task fits in one output and a stranger would need briefing to do it, use the five lines. Below that, just ask. Above that, expand.
Why this beats memorizing tricks
You may have seen long lists of prompt engineering techniques — and many of them are genuinely useful. But techniques are answers to specific problems, and most people's problem isn't exotic. It's that the request was missing basic information. Role, context, goal, constraints, format: these five lines cover roughly the same ground a good creative brief, a good support ticket, or a good delegation message covers. That's not a coincidence. Clear requests have looked like this since long before AI.
The other advantage is speed of habit. One template, reused daily, becomes automatic in a week or two. You stop staring at the empty box, you stop writing "make it good please," and you start noticing — often before you even send the brief — which line you can't fill in. That last part is sneaky-valuable: if you can't write the goal line, the AI was never the bottleneck. You hadn't decided what you wanted yet.
Practice it on today's challenge
Reading about briefs is like reading about push-ups — pleasant, low-impact, and not the thing that builds the muscle. The template sticks when you use it on real scenarios and get feedback on what you missed.
That's exactly what today's challenge is for: one realistic scenario a day, you write the brief, and the AI coach scores it from 1 to 5 stars with specific feedback — usually pointing at one of these five lines. It takes a few minutes, it's free, and after a couple of weeks the five questions start asking themselves.
Bring the template. See which line you forget first. Most people are surprised — and most people fix it within a week.