Chapter 4 · Structured workflow · 15–25 minutes

Be specific, be honest, be demanding

What this is

Chapter 4 makes the case that the difference between a useless AI response and a useful one is not prompt engineering — it’s writing a proper brief and then pushing AI to do work it won’t do on its own. You saw Gemma do this in the book: she walked AI through a structured brief for her Marketing Director cover letter, named her weakness honestly (no B-Corp experience), and asked AI to interview her before drafting. The result was a letter that no one-line prompt could have produced.

In the next fifteen to twenty-five minutes you’re going to do the same on something real you need to write or decide this week. Not a test. A real thing.

What you’ll need

15 minutes for the core exercise, or up to 25 if you do the optional second move
A real task you need to produce this week: a piece of writing, a decision to think through, a conversation to prepare for, a document to draft
Your usual AI tool open in a new conversation

If you don’t have a real task this week, bookmark this page for when you do — the exercise needs real material to teach what it’s supposed to.

The brief format

The structured brief has five fields. The chapter argued for the first three (context, task, constraints); the last two (what good looks like, what I don’t want) are where most readers see the biggest lift.

Context — what’s happening, what you’ve tried, what constraints you’re working with. Be honest here. If you have a weakness or gap, name it. This is the chapter’s middle principle — be honest — and it’s the field most people fudge.
Task — what specifically you need help with
Constraints — operating limits on the output: length, tone, format, audience
What good looks like — the ideal output in one sentence
What I don’t want — specific things the output should avoid (the difference between this field and constraints: constraints are the envelope, this is the landmines)

That last field does more work than people expect. Telling AI what to avoid is often clearer than telling it what to aim at. “Don’t sound corporate, don’t use the word leverage” is more useful than “sound natural”.

Worked example — Gemma’s brief

Here’s Gemma’s situation from the chapter, translated into the five-field structure. Read it for the shape, then write your own.

Context: I’m applying for a Marketing Director role at a B-Corp consumer brand that sells sustainable household products. I’ve spent eight years in consumer goods marketing, the last three leading a team of six. My biggest achievement was repositioning a legacy cleaning brand as eco-friendly, which grew market share by 12% in two years. The job listing emphasises “authentic brand storytelling” and “community-led growth.” My weakness is that I’ve never worked at a B-Corp specifically.

Task: Help me write a cover letter that plays to my repositioning experience while being honest about the B-Corp gap.

Constraints: One page maximum. Lead with the repositioning story. Acknowledge the B-Corp gap once without dwelling on it.

What good looks like: A letter that makes the hiring manager curious enough to want to interview me — leads with my strongest evidence (the 12% growth), addresses my obvious gap without apologising for it, and ends with a specific point about why this particular role matters to me.

What I don’t want: Generic enthusiasm about sustainability. Lists of skills. Anything that sounds like it could have been written for any B-Corp marketing role. Don’t open with “I’m writing to apply for…”

Notice three things Gemma did that make this work. She named her weakness directly (the B-Corp gap) instead of hoping AI wouldn’t notice — that’s the be honest principle in practice. She gave a specific output target (curious enough to interview) instead of “a strong letter”. And the “what I don’t want” field is concrete — generic enthusiasm, lists of skills, interchangeable phrasing — not abstract.

Your turn

Pick your task. Copy this blank template and fill it in. Don’t rush the fields — most of the value of the exercise is in writing the brief, not in reading what AI sends back.

Context: [Describe your situation — what’s happening, what you’ve tried, what constraints you’re working with. If you have a weakness or gap, name it here. Don’t hide it.]

Task: [What specifically do you need help with?]

Constraints: [Operating limits on the output — length, tone, format, audience]

What good looks like: [The ideal output in one sentence. Be specific about what makes it good.]

What I don’t want: [Specific things the output should avoid — be concrete, not abstract.]

Paste it into your AI tool and read what comes back. Don’t accept it yet.

The interview move

This is the move Gemma made next in the chapter, and it’s the strongest single follow-up you can run. Even a good brief leaves things out. Asking AI to interview you surfaces what you didn’t think to include — and it’s the chapter’s be demanding principle in practice.

Before you go further, interview me first. Ask me at least five questions about this — things you’d want to know to make the next draft materially better. Things I might not have thought to tell you.

Answer the questions honestly, including the ones you don’t have a great answer for. (Honest “I don’t know” is more useful than a polished guess.) Then ask AI for the next draft. The difference between this version and the first one is usually substantial.

Optional — find what’s missing

This is a second move, and it costs about ten more minutes. It works best after the interview, not instead of it. Skip it if you’re short on time; the core exercise is complete without it.

Now tell me: what are the two strongest objections a sceptical reader would have after reading this, and how should I address them? Is there anything in the current draft that sounds good to me but might land badly with someone else?

This catches what you and the interview didn’t surface. AI will either spot a real weakness you missed, or invent a problem that isn’t actually there. Your job is to judge which it just did. That judgement — knowing when AI is right and when it’s overstating — is one of the most useful skills the book teaches.

What you’ll notice

If you write a real brief on a real task and run the interview move, two things usually happen:

The first response is better than what you’d have got from a one-line request — but not transformatively better. The brief format alone gets you a good first draft; iteration gets you a final one.

The interview move surfaces something you hadn’t planned to include. Usually it’s the most useful part of the exercise, because it tells you what’s actually in your head that hasn’t made it onto the page.

If you also ran the optional second move, it will either surface a blind spot you missed, or hallucinate a problem that doesn’t exist. Your job is to judge which it just did.

The brief format is reusable. Save the template, adapt the fields to your situation, and you have a starting structure for every meaningful AI conversation. You now have a template you can reuse, and a finished draft you can send today.