TL;DR: Most failed AI pilots are leadership problems in disguise. The real moat isn't which model you use, it's the people, the system, and the story clarity you give AI to work with. Get those right first.
I walked into my last CMO role and found a team that had been running without a marketing plan, without a shared understanding of who the customer was, and without much of a way of working together. The culture within the company was becoming more and more oriented towards becoming "AI-native" but without a strategy. It was up to the leaders. But I knew inside that anything related to AI would be a fools' errand because the fundamentals listed above weren't nailed down.
I couldn't just say "everyone start using Claude" or "let's build some workflows" because that kind of jump assumes the team already knows how to work together. Mine didn't. They didn't have clear job descriptions. Ownership was fragmented. Nobody could agree on who owned what, let alone how to brief an AI tool on it.
So before I touched a single AI tool, I did something much less sexy: I got the team clear on how they actually work.
The unsexy thing first
I introduced what I call the GSD model: Get Shit Done. It's something we lived by at Shopify, it was codified in our teams. It wasn't necessarily something we had to learn or teach. Rather, it was a mental model. In essence, it was another way of saying — have an idea, bring smart people together, build a plan, and get shit done!
The concept isn't complicated. It's just a documented way of working: how does insight enter the system, how does it turn into an idea, how does the idea become a brief, how does the brief become something that ships? Insight → Ideate → Create → Launch. That's it. But you'd be shocked how many marketing teams have never written this down, let alone agreed on it.
The other thing about the GSD model: you don't hand it down from on high. You co-create it. Your team needs to live it, push back on it, reshape it. That's what makes it stick versus becoming another doc nobody reads.
So job one wasn't AI. It was getting people clear. Only once that was in place could I honestly ask: what's our story, and who are we telling it to? And only once that was in place did AI start to actually mean something.
The leadership crisis nobody's talking about
Here's something the AI hype cycle doesn't want to say out loud but I'm happy to shout it from the roof tops: most failed AI pilots are leadership problems in disguise. Unclear objectives. Five versions of the customer floating across different decks. Roles that don't map to the actual work. A culture where people have learned to do what they want rather than what the business needs — not out of malice, but because nobody ever clearly said otherwise.
AI is just another new, super powerful technology. This conversation is about transformation. Transformation is done via people. These problems predate AI by years. AI doesn't create them. It accelerates them.
If your foundations are solid, AI amplifies that. If they're not — if nobody can agree on who the customer is, or what "good" looks like — then AI helps you scale the confusion faster. Or, your people feel inert because they don't know what step to take next.
When someone tells me their AI pilots aren't working, my first questions are never about the technology. They're:
- Does your leadership team agree on who the customer is?
- Do your people have clear roles?
- Is your briefing process creating clarity or rework?
- Do your people feel safe enough to experiment or are they afraid?
If the humans can't agree, the AI definitely won't fix it.
Your people are the moat. AI is the lever.
In an era where everyone has access to the same models and tools, your competitive advantage isn't the technology. It's what you bring to the technology.
That means:
→ Your people: their judgment, their taste, their deep knowledge of your customer.
→ Your system: how you brief, how you review, how you close the loop from customer insight back into the story.
→ Your story: the clarity and coherence of who you're for, what you promise, what you can prove.
Think about what that looks like in practice. A marketer opens a campaign brief and the right ICP definition is already there. The current proof points are already there. The approved narrative is already there. The first AI draft is 70% right not because the model is smarter, but because the inputs are coherent.
Without that? They're pasting a soup of old slides into ChatGPT and hoping something useful comes out. It will produce something. But it's stitching together inconsistent inputs, because nobody ever sat down and decided what the truth actually was.
The companies that win with AI won't be the ones who moved fastest. They'll be the ones whose leaders did the unsexy upstream work first. Fundamentals first. Systems second. Story third. AI fourth.
Most of the industry has that completely backwards. Let's flip it.
Read Part 2: How I'm Doing AI as a CMO…What do you actually build?
Anne-Marie Goulet is the founder of Swim Club, a strategic practice for founders and marketing leaders who are done with the theatre. She spent 15 years inside Salesforce, Shopify, and WordPress Enterprise leading global marketing teams. Say hi at anne-marie@joinswimclub.com.
