The missing layer
I recently called it vibe marketing — the way most marketing actually runs, beneath the frameworks and the processes. But naming a problem isn't explaining it. So here's the explanation.
Vibe marketing isn't a bad habit you can coach away. It's not about lazy people or bad briefs. It's an infrastructure problem: marketing has no artifact that captures the thinking behind the work. It has plenty of artifacts that describe what to make — briefs, brand books, message houses, playbooks. None of them answer a different question entirely: what must be true for this to work?
That question has never had a home in marketing. It needs one.
A typical Tuesday
Here's how marketing actually works in most organizations. A mid-senior marketer at a multi-product company sits down to create something. Here's what's in front of them:
A job description that says "drive awareness and adoption among enterprise developers." Direction without mechanism. A brand book with visual guidelines and tone words — professional but approachable, innovative yet trustworthy. It tells you what the brand looks like. It says nothing about what the audience currently believes or what needs to change. A set of playbooks that tell you how to run the process. And Slack — the actual knowledge transfer system.
So you read the brand book. You skim last quarter's campaign deck. You piece together partial context from two Slack conversations and a doc someone shared three months ago. You fill the remaining gaps with your best professional judgment. You produce something.
It goes to review. The reviewer — who holds different context in their head, assembled from their own set of conversations and documents — says "this doesn't feel right." They can't say what specifically is wrong, because there's no shared document that defines what "right" would look like. The feedback is taste. Back to revision. Another round. Maybe a third. Eventually someone senior enough says it's fine (because it's late), and it ships.
The result is a patchwork of individual gap-filling presented as a coherent piece. Without a spec, the only standard is whether the output looks like marketing. And everything looks like marketing. A social post looks like a social post. An email looks like an email. It compiles because it reads like text. But does it do what it needs to do? You can't say — because there's nothing to check it against except taste.
Why this wasn't always a problem
This used to be fine. I started my career at BBDO, working in creative couples — copywriter and art director, together all day, every day, with a strategist close enough to overhear your arguments. The brief could be vague because context traveled through proximity. You didn't need to write down the audience's emotional state — your strategist friend told you between cigarettes. You didn't need channel adaptation specs — the channels were TV, print, and outdoor, which was basically print with a bigger font. Three people could hold the complete knowledge surface. The informal system worked because the scope fit inside a few heads.
That world is gone. It broke in two directions at once.
Horizontally, channels exploded. Digital, social, programmatic, influencer, email sequences, SEO, community, developer relations — each with its own norms, formats, and audience behaviors. No creative couple can hold all of that. Knowledge got distributed across specialists, and the moment it distributed, proximity stopped being a transfer mechanism.
Vertically, organizations got complex. Multiple products, segments, functions, markets. At Nike, I managed marketing across membership, retail, brand, and sport categories — each with their own teams, their own understanding of the consumer, their own way of translating strategy into execution. Team A discovers something real about an audience segment. Team B, different product line, same audience, figures out the same thing six months later. Team C never learns it at all. The insight sits in Team A's Keynote, in a folder nobody outside the team will open.
Both directions broke the same thing: the ability of human proximity to replace written-down thinking. But nothing formal took its place. The architecture was designed for three people in a room with three channels. That world ended fifteen years ago. The architecture never updated.
What's actually broken
The root cause is specific: marketing has no knowledge artifact.
Other disciplines write down why something should work before building it. Not because their artifacts are always correct — plenty of terrible specs exist — but because they're specific enough to be wrong in an identifiable way. A bad spec is still debuggable. You can point to the part that's wrong. A bad vibe is just... a bad vibe.
Marketing has execution artifacts (brand books, style guides, templates) and process artifacts (playbooks, approval workflows, calendars). It has nothing that captures the reasoning. The decisions, the assumptions, the conventions, and critically, the boundary between them. Your brand color is a decision. Your choice of emotional imagery over technical diagrams is an assumption. The difference matters enormously when something underperforms — you don't debug the brand color, but you might need to debug the emotional imagery. Right now, both sit at the same level in the brand book. Both are just "how we do things."
This missing artifact produces three specific failures.
The thinking never assembles. The knowledge needed for good marketing exists — it's just scattered across people, channels, and documents that were never designed to come together. Every executor independently reconstructs a partial version from whatever they can access. Without a shared contract for what the work must achieve, everyone stays dependent on everyone else. You need the strategist in the room, the brand person on the call, the channel expert in the thread — not because collaboration is bad, but because the context lives only in their heads. A spec is a boundary that enables parallel work. Without it, coordination overhead scales with every person you add.
Knowledge doesn't compound. Strategic knowledge — what you learned about audiences, what arguments work and why, which mechanisms produce which shifts — evaporates. It never gets externalized in a form that can travel across teams or persist across time. Operational knowledge compounds fine: brand guidelines accumulate, approval processes mature, templates improve. Strategic knowledge resets with every campaign. People leave, teams rotate, agencies change. The new people produce work that looks right — follows the brand book, matches the tone, hits the format requirements. But the original strategic intent is gone. Why this audience? Why this angle? What assumption was behind that choice? Nobody knows, because it was never written down. The surface stays professional while the foundation erodes underneath. Nobody notices until results drift and nobody can explain why.
You can't debug what you can't see. Something underperforms. The team sits in a room. "The messaging didn't land." "Maybe the channel was wrong." "I think the audience was off." These are all plausible, and nobody can say which one actually broke — because the assumptions were never made explicit enough to test. You can't do a post-mortem on vibes. You can only have opinions. So the team adjusts everything a little, learns nothing structural, and the same failure pattern shows up again in three months.
A spec shifts discovery to before execution begins, where mistakes are cheap to fix. You catch "we're targeting the wrong audience state" at spec review, not when the campaign underperforms. You catch "this argument assumes a problem the audience doesn't feel" before a creative team spends two weeks building around it.
(And if you're thinking "but we A/B test everything" — A/B testing debugs outputs. This headline versus that headline. It tells you A beat B. It doesn't tell you why. Was it the audience assumption? The value proposition angle? The emotional register? The optimization is tactical. The learning isn't structural.)
The AI moment
This is where it gets urgent.
AI is the most literal executor you've ever worked with. It does exactly what the input says. Every gap in your brief — audience, intent, structure, reasoning — it fills with statistical averages. The result reads well. It sounds professional. It follows the tone. But reading well isn't working well. Plausibility is not correctness.
Human talent masked this for decades. A skilled copywriter receiving a vague brief would fill every gap with professional judgment, audience intuition, and craft. The output was good despite the brief, not because of it. AI doesn't fill gaps with judgment. It fills them with the average of everything it's seen. The vagueness that human talent absorbed is now reflected back, in full, as mediocre output.
Here's the thing though: give AI a real spec — with a defined audience state, a clear mechanism, explicit assumptions — and the output changes dramatically. AI executes well against a contract. The problem was never the tool. The problem is that the contract doesn't exist. And that means the real work isn't producing content. It's defining what the content must do.
AI didn't create the problem. It made it impossible to ignore.
The real point
But here's the thing I keep coming back to. This isn't really about AI.
The spec isn't a better brief. It's a shift in what counts as "the work." Right now, marketing treats the content as the first truth — the thing that gets reviewed, debated, revised, approved. The thinking behind it stays invisible, reconstructed differently by everyone who touches it.
The spec inverts this. The thinking is the first truth. The content is an implementation of it. You review the thinking first. The execution then has something real to be measured against — not taste, not "does this feel right," but does it fulfill the contract. Does it do what we explicitly said it needed to do, for the audience we explicitly described, through the mechanism we explicitly chose.
AI made this urgent. But we owed it to ourselves all along.