Most agency teams are drowning in AI output that doesn’t sound like their clients.

You prompt the tool. It returns something that technically checks the boxes—right length, right format, right topic. But somewhere in the polishing process, your account director squints at the screen and says: “This doesn’t sound like them.”

Then comes the editing tax. Forty minutes rewriting a draft to recover the brand’s actual voice. Do that five times a week across six clients, and you’ve lost everything the AI was supposed to save.

The problem isn’t the tool. It’s the brief.

A prompt is not a brief. A prompt is a command to a machine. A brief is the entire context a human needs to make a decision. Most agencies treat them as interchangeable. That’s why the output fails.

Here’s what actually works: the 4-layer brief.

Layer 1: Voice

Who does this brand actually sound like?

This is where most briefs collapse. They describe the voice with adjectives: “professional,” “friendly,” “authoritative,” “warm.” An AI trained on ten million documents has already heard those words. It will produce professional-sounding words. But it won’t produce this brand’s professional voice.

Instead, use specific comparisons.

“Sounds like our founder when she’s made a decision, not when she’s second-guessing.”

“Think Elon’s Twitter posts from 2015—direct, a little combative, no softening language.”

“Our customer is tired of wellness brands trying to be her friend. She wants a doctor who respects her time. Write like one.”

Concrete beats abstract every time. An AI can’t do much with “be authentic.” It can absolutely work with “no exclamation marks, ever” or “the tone is ‘figured it out the hard way’ not ‘learned this in a masterclass.’”

Layer 2: Constraint

What will this brand never say?

This layer is more useful than any list of “approved words.” Constraints are about what’s forbidden—the taboo topics, the words that trigger legal, the claims the brand won’t make even if they’re true.

“No promises we’ll be ready before 2027.”

“Don’t mention our Series A funding—makes us sound like we’re hiring.”

“Never position this as a solution to their biggest problem. Position it as a solution to the problem they didn’t know they had yet.”

“No ‘we’re passionate about.’ No ‘we believe in.’ No founder philosophy.”

Constraints are clarifying. They shrink the possible output space. They make it harder for generic AI to produce generic work.

A junior copywriter who knows the constraint won’t waste time writing around it. Neither will the AI. The brief gets sharper. The output follows.

Layer 3: Intent

What should the reader feel, do, or remember?

Write this in one sentence.

“They finish reading and understand we’re cheaper than hiring a junior copywriter full-time.”

“They realize their current setup is leaving money on the table.”

“They save the post because it explains something they’ve been trying to articulate.”

If you can’t write the intent in one sentence, the campaign isn’t ready for the AI. Send it back to strategy. When intent is vague, output is vague.

Intent tells the AI which details matter and which are window dressing. It’s the difference between “write a post about our new feature” and “write a post that makes a midsize agency director realize they’ve been underselling their services.”

Layer 4: Artifact

Format, length, platform, deadline.

Specific.

Not: “social post”

Actually: “LinkedIn post, 250 words max, video thumbnail 1200×628, drops Wednesday 8am GST”

Not: “email copy”

Actually: “Retention email to users who generated fewer than 5 posts in 60 days, 3 paragraphs, max 150 words, CTA is ‘Start a free workshop,’ from the CEO email address”

Artifacts constrain the creative problem in practical ways. They tell the AI what “done” looks like. They’re the difference between a rambling 800-word draft and a focused 120-word post.

Where Most Teams Get It Wrong

Most briefs that land on an AI team’s desk have layers 3 and 4—intent and artifact. Often just barely.

Layers 1 and 2—voice and constraint—are missing or vague. Sometimes they’re left to “general knowledge” of the brand. The AI doesn’t have that knowledge. The junior copywriter who’s bouncing between clients doesn’t either.

So the output is technically correct, strategically on-target, and generically executed. Every revision pulls it back toward this brand’s actual voice. Every revision reveals something the brief should have said from the start.

Multiply that across a quarter, and the editing tax has compounded into a revenue leak.

Try It This Week

Pick one client brief that went out this week. Rewrite it with the 4 layers.

Don’t change the artifact. Don’t change the intent. Just add: one specific voice comparison and three hard constraints.

Run it through whatever tool you’re using. Compare the output to what you actually produced.

The difference won’t be subtle.

You don’t need new software to see this work. You need a different brief.