“Can we just automate all of Marketing?”
That’s what a CMO was recently asked by the C-suite.
It’s a rhetorical question dressed up as a strategic one. And it reveals exactly the problem Mark Stouse and I unpacked in our recent Causal CMO chat.
The problem is the question sounds like it’s just about Marketing.
It isn’t.
Because if AI can replace the logic of Marketing, it can replace the logic of any function. Including the C-suite. AI does not respect org charts.
The only real moat is the quality of your thinking. That’s the decision layer.
Here’s the recap.
The research Mark and his team have done over the past two years shows that improving your prompt architecture produces a 5x to 10x improvement in output quality.
But this isn’t about writing better prompts. Prompts are just an expression of how you think. If the thinking is shallow, the output will be too. Faster, but shallow.
“The greatest test that AI poses to all of us is our ability, and our willingness, and the time that we devote to improving our own thinking. Our own logic. [Our] ability to think it through, to tell the machines what they need to be doing.”
Classes in logic and rhetoric were quietly dropped from business education roughly 70 years ago. Deemed unnecessary for middle management. That decision is coming due now.
AI only knows what you give it.
“An AI agent is a pale shadow of you.”
That’s the difference between AI and HI.
AI can return answers. It can challenge your wording. It can even ask follow-up questions.
But in a Socratic exchange, AI’s questions are derivative. Human questions can be original.
That matters in GTM. The team’s value is not just using AI to answer faster. It’s asking the questions the tool would not know to ask.
Which means the decision layer (the logic that sits above your tools, dashboards, data, and automation) is the only thing AI can’t supply on your behalf.
This is why starting with decisions, not data, is so important. It’s all about the quality of the thinking before getting into the tactics.
When Mark’s firm takes on a new client, the first models they build have nothing to do with the client.
“The very first set of models that we do for them has nothing to do with them. It has everything to do with really understanding the marketplace in which they are operating.”
That output gets shared across Sales, Marketing, Product, Customer Success, and the C-suite: quantified headwinds, tailwinds, crosswinds. One shared version of “today’s reality” (that’s key) before anyone opens a dashboard.
Without it, Sales has its version of the market, Marketing has another, and Product lives in a third. Calling that a coordination problem misses the point. Each function thinks it’s right. A causal model puts the actual marketplace in the room, and the results are usually a jaw-dropper.
In a headwind environment, effectiveness craters even when the work is solid. Just like a plane has to burn more fuel, headwinds drain the energy out of everything.
“In most cases, the go-to-market teams have peanutbuttered their efforts across so many different things that even with additional spend, there’s not enough juice to get over the top of what’s opposing you.”
That’s why CAC keeps spiking even when teams spend more. Seventy to eighty percent of what drives any GTM outcome sits outside what your team controls. If you’re not modelling it, you can’t explain the results to yourselves, let alone to the board.
Left to its own devices, AI confirms your assumptions and sends you away feeling validated about a plan that hasn’t been stressed. It’s the ultimate Yes Man.
The fix is simple: make it argue.
“Gen AI is hardwired to please you. But it can’t define that. Only you can define that.”
Mark’s “5 AI Peer Review System” loads the same question through multiple AI tools. It sets them against each other. They recognise each other’s digital syntax. What follows is, in his words, “ferocious one-upmanship.” That’s where the useful friction starts.
“AI is not really replacing human intelligence. It is showing us the holes in our intelligence.”
Gen AI’s strengths are asymmetrical. It can’t consciously give you the right answer. But it’s very good at poking holes in what you assume.
The questions to drive toward:
Feed those takeaways into a causal model to understand what’s actually driving the outcome.
And by the way, correlation still plays a role in causal AI. But it’s used for discovery, not as the final readout.
When Mark’s team ran a cross-assumption analysis for a client, loading proposals from Sales, Marketing, Product, Customer Success, and Professional Services into a causal model, something became obvious fast.
Every function had positioned itself as the MVP. I mean, why not? Marketing implied it was most important. Sales said the same. Product lived in its lane.
“In fact, the only reason why a Marketing strategy or a Sales strategy even exists is that there’s not a strong Business strategy. So in the absence of that, people make up other strategies.”
A Marketing strategy without a Business strategy above it is a resource allocation argument dressed up as a plan. Same for Sales. Same for Product. Everyone’s protecting budget and turf because nobody’s drawn the map above them.
When the causal model mapped the network of causes and effects, that became hard to argue with. Sales can’t hit its number without Marketing because Marketing is a non-linear multiplier of Sales productivity. CS protects what Sales closes. And all of it sits inside a marketplace none of them control.
The aha moment? Nobody had a dinghy. They were all in the same boat.
Putting efficiency ahead of effectiveness is the core failure mode of AI adoption right now. Teams are using AI to do more of what they were already doing. But if the underlying motion wasn’t working, speed makes it worse.
“You’ve got to make sure that what you’re doing is effective. That is the first thing before you try to scale it, make it more efficient, all that kind of stuff. Otherwise, you’re just driving faster and faster in the wrong direction.”
The adoption data points in the same direction. McKinsey’s 2025 State of AI report shows 88% of companies now use AI in at least one business function, but only 39% report enterprise-level profit impact. Gartner predicts more than 40% of agentic AI projects will be cancelled by the end of 2027 due to rising costs, unclear business value, or weak risk controls.
Usage is outrunning decision quality. By a lot.
The billions wasted in GTM over the last 20 years weren’t wasted because teams were stupid. They kept doing the same things faster.
“Whatever you’re doing, you’re fighting the last war. And you’re losing the war that’s actually happening in real life as a result of that.”
AI doesn’t fix that. Better thinking does.
Three things worth doing before greenlighting the next AI project:
Missed the session? Watch it here.
If you like this content, here are some more ways I can help:
Cheers!
This article is AC-A and published on LinkedIn. Join the conversation!
