Getting a usable first draft out of a blank brief

Someone asks for “something for the newsletter” or “a post about the new feature.” You paste that straight into AI and get back three paragraphs of confident nothing: generic, on-topic, and useless. The instinct is that the tool wrote badly. It didn’t. It wrote to the brief it was given, and the brief was one sentence.

A blank brief in produces a blank draft out, dressed up in full sentences.

---

The five questions

Before writing anything, answer these. Two minutes, not a meeting.

1. Who reads this, specifically? Not "our audience". The person, the role, what they already know.
2. What should they do, think, or feel after reading it? One outcome, not three.
3. What do they already know that we don't need to re-explain?
4. What's the one fact, number, or detail that makes this worth their time?
5. What's it replacing, or what happens if we skip it entirely?

Question 5 is the one people skip and the one that matters most. If nothing happens when a piece doesn’t run, that’s worth knowing before you spend an hour on it.

---

Turn the answers into a prompt

Don’t hand the five answers over as a list and ask for a draft. Ask for the brief to be checked first.

Here's what I know about this piece:
[paste your five answers]

Before drafting: what's missing here that you'd need to write this well?
List the gaps, don't guess and fill them in yourself.

This catches thin answers before they become a thin draft. “Who reads this” answered as “our customers” is a gap, not an answer, and it’s cheaper to notice now than after a full draft comes back generic anyway.

---

Now ask for the draft, with the limits stated

Using the brief above, write [format, e.g. a 150-word newsletter blurb].

Rules:
- Lead with the fact from question 4, not a general opener
- Don't restate what the reader already knows from question 3
- One clear next step, matching the outcome in question 2
- No "unlock", "seamless", "game-changing", or similar. Plain language only

Stating what not to do earns more than stating what to do. Left alone, most drafts default to a warm generic opener and a vague call to action, because that’s the average of everything the pattern has seen, and your actual reader isn’t average.

---

Edit for the thing a brief can’t fix

A good brief produces a draft that’s specific instead of generic. It doesn’t produce one that sounds like your brand, or one that’s true. Read the draft for two things a prompt can’t guarantee:

Does it sound like a person wrote it, or like the average of everyone who’s ever written a newsletter? If every sentence could run under a different company’s name unchanged, the brief answers were probably still a little too generic.

Is every claim in it something you can actually stand behind? A confident sentence about “our most popular feature” or “customers love this” needs a number behind it or it comes out. See the honesty rule: no benefit claim without something to show for it.

---

Worked example

The ask, as it arrived: “post about the new export feature.”

Five-question brief: read by existing users who already export manually every week; should make them try the new one-click version this week; they already know the product, don’t re-explain what the tool does; the fact that matters is it cuts a five-step process to one click; if skipped, the feature quietly launches to zero awareness and adoption stays flat.

Gap the check-first prompt caught: no stated tone, and no channel, a one-click headline for social reads very differently from three sentences in an email. Added both before drafting.

First draft needed editing for one line that claimed the feature “saves hours,” a number nobody had actually measured. Replaced with the concrete fact: five steps to one.

---

When it breaks

The draft is specific but wrong. It invented a detail that wasn’t in your brief, usually because a gap from the check-first step went unanswered and it filled the hole itself. Go back to the five questions and see which one was thin.

It ignores the “don’t do this” list. Long, list-heavy prompts sometimes lose the back half. Put the rules closer to the request, or ask again with just the rules restated.

Every draft still sounds the same regardless of the brief. That’s a tone problem, not a brief problem. Paste in two or three pieces you’ve actually published and ask it to match that voice specifically, rather than describing the voice in the abstract.

The brief took longer than the draft would have by hand. Some pieces are genuinely too short or too one-off to be worth this. A single internal Slack update doesn’t need five questions. Save the method for anything that’ll be read by people outside your own team.

You answered all five questions and still don’t know what this piece is for. That’s a real finding, not a prompting failure. A piece nobody can state the purpose of is usually a piece that shouldn’t get written yet.

---

What to try next

Save your five answers alongside the final piece, not just the draft. The next similar request, a different feature, a different month’s newsletter, starts from a filled-in template instead of a blank line, and that’s most of the time this method saves.