Stop outsourcing your marketing intelligence to AI. Do this instead.
How to turn your team's marketing intelligence into your most valuable asset, one that compounds with every marketing campaign and doesn't leave when your best people do.
Satya Nadella published something last week that speaks to how a modern marketing team will win once all teams are rebuilt around AI.
The title of the post was “A frontier without an ecosystem is not stable.” It sounds like something written for a CTO.
But it’s as applicable for anyone working in marketing.
Buried in the post is the clearest description I’ve seen of where the real leverage sits when you’re building a marketing team around AI. It’s not what models you use. It’s not the AI tools you buy. It’s something much more fundamental than that.
It’s the following loop, ‘The Marketing Intelligence Loop’
It’s how you build genuinely differentiated marketing. And real leverage against competitors who’ve outsourced their entire strategy to AI models, models that don’t just advise them, they execute for them, producing the same output for everyone who asks.
Node 1: Marketing Judgment
The most valuable input
Marketing judgment can’t be replaced by prompt engineering.
It’s the ability to understand a customer, your market, the outcome you’re striving towards, and knowing what matters. To look at a multitude of options on how to garner attention with your audience and know which one breaks through the noise. To make tough calls on what work to focus on and what work is merely a distraction. To decide what learnings from last month are worth capturing.
AI doesn’t have this. The AI model you’re using doesn’t have this. It has the marketing knowledge averaged out across an entire set of marketing tutorials. But it doesn’t have what makes your judgment better than others.
That’s why you can give two marketing leaders identical tools, identical models, and identical budgets, and they’ll produce radically different outcomes based entirely on the quality of their judgment.
Judgment can’t be derived from a prompt; it’s earned through practicing the craft.
Node 2: Your Intelligence Layer
How marketing intelligence used to walk out the door
David Ogilvy was obsessive about writing his judgment down. Every hard-won lesson about what makes advertising work, what headlines engaged, how to brief a creative team, what separates a good campaign from a great one, went into memos that circulated across every Ogilvy & Mather office. New hires didn’t just learn from Ogilvy. They read Ogilvy. When he retired to his château in France in the 1970s, the agency didn’t lose its standards. Because he’d spent 25 years capturing his judgment in writing, his intelligence stayed in the firm long after he left.
Ogilvy was exceptional because he was brilliant, and he wrote everything down.
That’s the bet every CMO should make with AI.
Here’s what Satya Nadella actually said that matters for marketers.
He warned that if companies feed their data, workflows, and expertise into a handful of AI models, those models absorb it and turn it into a generic service available to everyone. Your competitive edge gets commoditized. His framing: “a world where every company in every industry is ceding value to a few all-consuming models.”
The alternative is: build a proprietary intelligence layer. A compounding system where your learnings, your audience knowledge, your positioning decisions, your historical performance, all of it lives in something you own, not something a vendor controls.
For marketing, this means every campaign cycle adds to the layer. What worked with your ICP this quarter. Which message outperformed in enterprise versus mid-market. What your audience is worried about right now. The judgment calls your best people made, and why.
This is fundamentally different from a wiki or a shared Notion doc. Those decay because nobody has time to maintain them. The intelligence layer is fed from the work you do.
Nadella gives a test he calls digital sovereignty.
Can you swap out your AI model tomorrow and still keep your competitive advantage?
If your intelligence lives in the model, the answer is no. If it lives in your layer, the model is just an engine you plug in. You could move from Claude to GPT-6 to whatever ships next year, and your institutional knowledge would still be there, getting smarter.
Own the layer. Rent the model.
I wrote about my own version of this in The AI Second Brain — a personal intelligence layer built in Obsidian and Claude Code. The Context Layer in How to Rebuild Your Marketing Team is the same concept at org scale. The architecture is identical. The difference is whether it lives in one person’s workflow or across an entire marketing function. I’m building the org-scale version at HubSpot now.
As a note, I’m working on an updated and improved version of The AI Second Brain that’s portable, across devices, across any model you want to use, and can be easily installed for your use. Stay tuned.
Node 3: AI Models
Pluggable execution, rent, don’t own
Nadella wrote: “Without human direction, you have compute running in circles.”
This is what most tool evaluations miss. The model is not the strategy. It’s the engine that reads from your intelligence layer and executes at scale. It’s powerful, but only as powerful as the layer beneath it and the judgment that directs it.
Today it’s Claude. In 18 months, it’ll be something better. In three years, the model landscape will look nothing like it does now. The models keep changing.
The teams that are building their AI strategy around usage, are outsourcing their marketing to AI models. They’re basing their entire strategy on a model that gives them the same ‘best practice’ as everyone else.
The ones building their intelligence layer are building something differentiated that compounds and creates real leverage to win.
The marketing intelligence loop is simple: the AI agents & assistants you use read from your layer, produce output grounded in your context rather than generic training data, and surface new insights back to marketing judgment.
Marketing intelligence loop = Real Leverage.
What this means for marketing teams
The conversation changes from “which AI model should we use?” to “what intelligence do we need to own?” How do we ensure that the depth of our marketing knowledge becomes our most prominent asset?
Part of the marketing leader’s job is to own the intelligence layer. It’s not a technical role; it’s a marketing judgment role. Who decides what goes in, what gets trusted, and what gets retired?
Your AI is only as good as what it reads from. Two marketing teams with the same model, without this layer, will get similar output. The marketing team who plugs the AI model into their intelligent layer, will get something differentiated, which is a result of that team’s core marketing intelligence. Better yet, it compounds over time.
See it in practice
The concepts above are straightforward. The harder question is: what does this actually look like when you build it?
I built a lightweight demo to illustrate how this could work. See a quick walkthrough I did here.
The point of the video is to show this concept, an intelligent layer, persistent memory, derived from the core knowledge across your marketing teams, their learnings, what works, and what doesn’t.
Marketing has always run on intelligence, about customers, about markets, about what works. The problem was always that the intelligence was fragile. It lived in people. It left with them.
The marketing intelligence loop solves that. Marketing judgment feeds this layer. The layer grounds AI execution. AI surfaces new insights for judgment. Every cycle, the system gets smarter.
The model you pick doesn’t matter.
The marketing intelligence loop you build does.
Until Next Time,
Happy AI’fying
Kieran