The brief is the bottleneck
Creators and AI can now produce almost anything quickly. What they can’t do is decide what’s worth making. That job belongs to the brief.
When an ad underperforms, the instinct is to blame the execution: the edit, the creator, the music, the thumbnail. Sometimes that’s fair. More often, when we trace a weak ad back to its origin, the problem started earlier. The brief asked for the wrong thing, asked for too many things, or didn’t really ask for anything at all.
Production has become fast and cheap. Creators can turn around content in days, and AI tools can generate variations in minutes. That makes the brief the slowest, most valuable part of the process. Everything downstream inherits its clarity, or its confusion.
What a weak brief looks like
Most weak briefs share a few traits:
- They describe the product, not the customer.
- They list every feature and benefit, hoping something sticks.
- They specify a vibe (“fun, authentic, high-energy”) instead of an idea.
- They don’t say what the ad is meant to test, so nobody learns anything when it runs.
The result is content that looks fine and says nothing in particular. It’s the hardest kind of ad to learn from, because there’s no clear hypothesis to prove or disprove.
What goes into ours
Our briefs are short, usually a single page. Each one answers the same questions.
Who is this for? Not a demographic, a situation. “Someone who has tried three versions of this product and been disappointed each time” is more useful than “women aged 25 to 44.”
What do they believe right now? The objection or assumption the ad has to move. This is often the most important line on the page.
What is the single job of this ad? One message. If the brief contains three, it’s really three briefs.
What’s the proof? A demonstration, a review, a comparison, a before-and-after, a number we can stand behind. A claim without proof is just an assertion.
What are the opening options? Three to five possible hooks for the first seconds, ideally in language lifted from real reviews and comments.
What format and length? And for which placements.
What are we testing? The hypothesis, in one sentence. “Leading with the problem will outperform leading with the product.” This is what turns an ad into a lesson.
Mandatories and no-gos. Claims we can’t make, brand rules, anything legal.
Briefing people and briefing models
The same brief works for both creators and AI tools, with different emphasis.
Creators need the why more than the what. Give a good creator the customer, the belief to shift and the proof, and they’ll find a more natural way to say it than any script. Over-scripting is one of the fastest ways to make real people look like actors.
AI tools need the what more than the why: explicit constraints, examples of what good looks like, and a narrow task. “Write ten hooks for this concept, under eight words each, in the style of these three reviews” produces far better results than “write some ad copy.”
Sources for better briefs
The best raw material for briefs is rarely inside the marketing team. We look at:
- Product reviews, especially the three-star ones, which tend to be the most specific.
- Customer service conversations and returns reasons.
- Comments on the brand’s own ads and on competitors’.
- Search terms customers actually use.
- The results log from previous tests.
AI is genuinely useful here. It can cluster hundreds of reviews into themes, pull out repeated phrases and surface objections we might miss. The judgement about which of those themes is worth an ad still sits with a person.
The habit that matters
If there’s one change worth making, it’s this: never brief an ad without writing down what you expect to learn from it. Over a few months, those sentences turn into the most valuable document a brand owns. It’s a record of what its customers believe, what moves them, and what doesn’t. Good creative follows from good questions, and the brief is where the questions get asked.