Marketing teams are spending months cranking out AI “skills” and prompt packs, only to ship content that feels clean but flat. I’ve seen it across crypto, social media, and SaaS: the output looks fine, yet nothing hits. Here’s my take after watching Marketing Against the Grain: skills aren’t the ceiling—context is. If your AI doesn’t know your audience, your voice, your position, and your buyer journey, it will average out the internet and hand you average work.
The Real Problem With AI Marketing
Kipp Bodnar and Kieran Flanagan make a sharp point using a Pixar story. Pixar didn’t fix movies by swapping directors or tools. They built a “brain trust”—a shared layer of context and feedback that lifted every film. That’s the missing layer in how most teams use AI today.
“The skills aren’t the problem. They never were. The problem is what the skills are actually building on top of.”
That’s the hard truth. You can stack prompt files and clever macros forever. Without a shared intelligence layer, each skill starts from zero. And zero gives you safe, samey work.
“What was missing was shared intelligence… a way for the collective wisdom to actually flow across every project.”
As a builder who’s launched products and media for decades, I agree. AI needs guardrails and nutrition before it needs more tricks. Feed it your living context, not just tasks.
What The Foundational Layer Looks Like
Their framework is refreshingly practical. It’s not another dusty brand deck. It’s a small set of markdown files that guide every skill, updated often and referenced only when needed. The result: sharper outputs, less fluff, and a voice that sounds like you.
- Audience Delight Profile: Real words your audience uses, what lights them up, what turns them off, and the phrases they never say.
- Creator Style: Your tone, patterns, do’s and don’ts, and how you open and close. Short, crisp, and specific.
- Positioning Map: Your claim, competitors’ claims, where you win, where it’s contested, and what white space you own.
- Customer Journey Intelligence: How people find you, what triggers action, their objections, conversion moments, and churn risks.
Each file declares when it should be used. Skills scan the folder and pull only what’s relevant. That stops the model from choking on extra context and keeps the system fast.
“Don’t obsess so much over the skills. Obsess over the context the skill is pulling in and make sure it pulls in the right context for the right task.”
My Playbook For Making It Work
I’ve built and sold digital products long enough to know this: consistency beats cleverness. Here’s how I’d roll this out in a week without blowing up your workflow.
- Start tiny: draft one-page versions of the four files. Keep each under 400 words.
- Train your AI to load only what’s needed. Content tasks load Voice + Audience; product pages also load Positioning; lifecycle emails pull Journey.
- Write five samples with and without the layer. If you can’t feel the difference, your files are too vague.
- Add a performance loop: tag outputs with campaign and channel, track CTR, signups, and retention. Update the files every quarter with what worked.
- Ban generic terms in your Voice file. Force specific words and phrases your market actually uses.
Some will argue better prompts fix this. They won’t. Prompts are instructions; context is memory. Without shared memory, your model forgets who you are every time.
Why This Matters Now
AI is flattening content faster than any trend I’ve seen. If you sound like everyone else, you’ve already lost. The fix isn’t more hacks. It’s a living context layer your whole team updates and your AI respects.
One more reason I back this approach: it compounds. Every improvement to your files boosts every skill, across email, ads, landing pages, and sales enablement. That’s leverage.
Final Thought
My opinion is simple: stop shipping skills that start from zero. Build your brain trust. Give your AI a shared context it can rely on, and refresh it with real performance data. This week, draft the four files, wire them into your skills, and run the A/B test. Then update them every quarter—no exceptions.
If you want standout work, feed your system standout context. Average data makes average content. Your audience—and your revenue—deserve better.
