Copied to clipboard!
Module 7 Level 1: AI Foundations for Design and Marketing beginner 1 hr

Capstone: Verify & Ship a Campaign Brief

Everything you learned, in one shipped brief.

Verify and Ship a Campaign Brief

This is where it all comes together. You take your Creative Context Document and one audience persona, build a real campaign brief, then draft a first-pass headline, subhead, and call to action. You ground every claim in something true, run the AI slop test, and log the human decisions you made so the work is yours to own. AI is the studio assistant that gets you off the blank page fast. You stay the art director who decides what ships.

Recommended tool

claude

Claude holds your context across a long thread and matches your voice well, which is what a campaign brief and first-draft copy need. If a claim must trace to a source document or a price sheet, read it in NotebookLM first so nothing gets invented, then write in Claude.

Bring to the exercise

  • Your Creative Context Document from Module 5
  • The audience persona you want to win (from Module 6, or a real target you know)
  • The one campaign you actually need to launch and its single goal
  • Any real, checkable facts or proof points you can stand behind
Step 1

Ground It in Your Context and Goal

Load the brand, the audience, and the single goal before any copy gets written.

What to substitute before pasting

  • [PASTE YOUR CREATIVE CONTEXT DOCUMENT] The one-pager you built in Module 5.
  • [PASTE THE PERSONA, OR DESCRIBE THE REAL TARGET] The audience persona from Module 6, or a real target you know well.
  • [THE SINGLE THING SUCCESS LOOKS LIKE, e.g. "50 newsletter signups in two weeks"] One measurable goal, not three.
We are building one campaign together. Use only what I give you and ask before you assume.

Here is my Creative Context Document:
[PASTE YOUR CREATIVE CONTEXT DOCUMENT]

Here is the audience persona I am trying to reach:
[PASTE THE PERSONA, OR DESCRIBE THE REAL TARGET]

The one goal of this campaign is: [THE SINGLE THING SUCCESS LOOKS LIKE, e.g. "50 newsletter signups in two weeks"].

Play back, in a few lines, who this is for, what they care about, and the one action we want. Do not draft anything yet.
What good output looks like
  • Restates the audience and the single goal in your terms.
  • Holds off on writing copy until you say go.

Verify before using AI's output

Step 2

Build the Campaign Brief

Turn the goal and audience into a one-page brief you would actually hand a designer.

Now write a one-page campaign brief from my context and goal. Use these sections:

- The audience and the one tension or problem we are speaking to
- The single message (one sentence a stranger would remember)
- The primary call to action
- Proof or reason to believe (leave blank where I have not given you a fact)
- Channels and format
- What success looks like

Use only what I have given you. Where a proof point is missing, write "NEEDS A REAL FACT" instead of inventing one.
What good output looks like
  • A tight one-page brief with one message and one CTA.
  • Missing proof is flagged "NEEDS A REAL FACT," never invented.

Verify before using AI's output

Step 3

Ground the Claims

Replace every "NEEDS A REAL FACT" with something true and checkable, or cut the claim.

What to substitute before pasting

  • [PASTE YOUR REAL FACTS, NUMBERS, OR PROOF POINTS] Things you can actually back up: real results, real features, real testimonials. No guesses.
Here are the real, checkable facts and proof points I can stand behind:
[PASTE YOUR REAL FACTS, NUMBERS, OR PROOF POINTS]

Fill in every "NEEDS A REAL FACT" using ONLY these. For each claim in the brief, tell me which fact it rests on. If a claim has no fact behind it, do not soften it, flag it as UNSUPPORTED so I can cut it or go find proof.

Do not add new claims. Do not round or improve my numbers.
What good output looks like
  • Every claim is tied to a fact you provided.
  • Anything without a fact is marked UNSUPPORTED, not smoothed over.
  • Your numbers are used exactly, not rounded or inflated.

Verify before using AI's output

Step 4

Draft the First-Pass Creative

Get a headline, subhead, and CTA on the page so you have something to direct, not a blank screen.

Now draft first-pass creative from the grounded brief. Give me three options, each with:
- HEADLINE (short, in my brand voice)
- SUBHEAD (one line that earns the click)
- CALL TO ACTION (the exact button or link text)

Follow my voice and obey my anti-references, nothing on my Do-Not list. Use only grounded claims. Keep it human and plain, no buzzwords. These are drafts for me to direct, not finals.
What good output looks like
  • Three distinct options, each with headline, subhead, and CTA.
  • On-brand voice, nothing from your Do-Not list.
  • No ungrounded claims sneak back in.

Verify before using AI's output

Step 5

Run the Slop Test and Pick a Direction

Pressure-test the drafts for generic, reflexive output before you fall in love with one.

Audit these three options like a skeptical creative director. For each one, answer:
- Could any brand have written this? (If yes, it is slop.)
- Does it match my brief and my brand voice?
- Is every claim grounded in a fact I gave you?
- Would the persona stop and care, or scroll past?

Rank them, tell me which is strongest and why, and show me one sharper version of the winner. Do not flatter me.
What good output looks like
  • Names which options are generic and why.
  • Ranks them against the brief, voice, and grounding, not vibes.
  • Offers one genuinely sharper version of the winner.

Verify before using AI's output

Step 6

Log the Human Process (manual)

Write down the decisions you made so the work is yours to own, not just AI output.

Help me write a short process log for this campaign, the record that shows the human work behind it. Walk me through filling in:

- The brief and goal I set
- The key prompts I used
- Every decision I made (which option, what I cut, what I rewrote and why)
- The facts I grounded the claims in and where they came from
- What I changed by hand at the end

Keep my original drafts and edits saved alongside this log. This is what makes the campaign mine to own, not the AI's. (Practical ownership habit, not legal advice.)
What good output looks like
  • A clear log of your decisions, edits, and the facts behind each claim.
  • A reminder to keep your drafts and edits saved.

Verify before using AI's output

Run the AI Slop Audit

AI gives you something that looks finished, sounds confident, and is completely forgettable. That is slop: generic, reflexive, could-be-any-brand output, sometimes with made-up claims baked in. Before anything you made with AI goes out the door, you run it through a fixed audit. You check whether it is generic, whether it matches your brief and brand, whether the claims are grounded, and whether it clears your accessibility floor. The AI helps you see the gaps; you stay the art director who decides what ships.

Recommended tool

claude

Claude is good at honest, structured critique against a brief and at separating grounded claims from confident guesses. If the piece cites a document or a number, check it in NotebookLM so the verification stays tied to the real source.

Bring to the exercise

  • One piece of AI-assisted work you are about to use (copy, a layout, a caption, an email)
  • Your Creative Context Document, including your anti-references and design principles
  • The brief or goal the piece was supposed to serve
  • The real facts behind any claims the piece makes
Step 1

Run the Slop Test

Find out fast whether the work is generic or actually yours.

What to substitute before pasting

  • [PASTE THE WORK, OR DESCRIBE THE LAYOUT] The actual copy, caption, or email, or a plain description of the visual layout.
  • [PASTE YOUR CREATIVE CONTEXT DOCUMENT] Your one-pager, so the audit judges against your real brand and anti-references.
Here is a piece of AI-assisted work I am about to use:
[PASTE THE WORK, OR DESCRIBE THE LAYOUT]

Here is my Creative Context Document:
[PASTE YOUR CREATIVE CONTEXT DOCUMENT]

Run the slop test. Answer honestly:
- Could any brand have produced this? Quote the lines or elements that feel generic or reflexive.
- Where does it lean on buzzwords, clichés, or filler?
- What is missing that would make it unmistakably mine?

Do not rewrite it yet. Just show me the slop.
What good output looks like
  • Quotes the specific generic or reflexive lines, not a vague verdict.
  • Names the buzzwords and filler.
  • Says what is missing that would make it yours.

Verify before using AI's output

Step 2

Check It Against the Brief and Brand

Confirm the work serves the goal and obeys your anti-references.

What to substitute before pasting

  • [PASTE THE BRIEF OR THE ONE GOAL] The brief or single goal the piece was meant to hit.
Here is the brief or goal this piece was supposed to serve:
[PASTE THE BRIEF OR THE ONE GOAL]

Check the work against it:
- Does it serve this goal, or wander off it?
- Does it match my brand voice?
- Does it break anything on my anti-reference Do-Not list? Quote any violation.
- Is the primary call to action present and clear?

List every mismatch as a specific fix I should make, not a silent rewrite.
What good output looks like
  • Says clearly whether the work serves the goal.
  • Quotes any anti-reference it breaks.
  • Confirms the CTA is present and clear.

Verify before using AI's output

Step 3

Verify the Claims Are Grounded

Separate what is true and checkable from what the AI confidently made up.

What to substitute before pasting

  • [PASTE YOUR REAL FACTS, OR WRITE "NONE" IF IT SHOULD MAKE NO CLAIMS] The proof you can actually back up. If the piece should make no claims at all, write NONE.
Here are the real facts behind any claims this piece makes:
[PASTE YOUR REAL FACTS, OR WRITE "NONE" IF IT SHOULD MAKE NO CLAIMS]

List every factual claim, number, or promise in the work. For each one, mark it:
- GROUNDED (matches a fact I gave you, name which)
- UNSUPPORTED (no fact behind it)
- WRONG (contradicts a fact I gave you)

Do not fix them. Just show me which claims I cannot stand behind so I can cut or correct them myself.
What good output looks like
  • Every claim is marked grounded, unsupported, or wrong.
  • Each grounded claim names the fact it rests on.
  • Nothing is quietly "fixed" for you.

Verify before using AI's output

Step 4

Clear the Accessibility Floor

Check the basics: readable contrast, alt text, plain language.

What to substitute before pasting

  • [PASTE YOUR ACCESSIBILITY FLOOR] The accessibility section of your Context Document.
Here are my accessibility needs from my Context Document:
[PASTE YOUR ACCESSIBILITY FLOOR]

Check this piece against that floor with a short, plain checklist:
- Is the language plain enough for my audience to read fast?
- Does any image need alt text that is missing?
- For a layout: is the text-to-background contrast readable, and is anything carried by color alone?
- Is the call to action easy to find and understand?

Give me a quick PASS / FIX on each, with the specific fix where it fails. This is a practical check, not a full compliance audit.
What good output looks like
  • A short PASS / FIX checklist against your floor.
  • Specific fixes where it fails, not generic advice.

Verify before using AI's output

Step 5

Ship or Send Back (manual)

Walk the manual gate. Nothing ships until it clears all four checks.

Walk this checklist by hand before this work goes out:

[ ] It is not slop, it could only be mine, not any brand's.
[ ] It serves the brief and breaks nothing on my Do-Not list.
[ ] Every claim is grounded in a fact I can stand behind.
[ ] It clears my accessibility floor.
[ ] I made the final creative call, not the AI.

If any box is unchecked, send it back for one more pass. If all are checked, it ships.
What good output looks like
  • You can check every box honestly.
  • The final call is yours, not the AI's.

Verify before using AI's output