How CMOs need to lead the AI shift
AI is reshaping how marketing gets done. Here is how to rebuild the operating model.
A lot of CEOs asked their CMOs to cut content marketers or designers, believing AI would compensate. Turns out, it didn't compensate. It scaled the mess. Whatever was already incomplete or complicated about how that team briefed, reviewed, and shipped work, AI did more of it, faster.
So here is what I’ve come to believe after a year of watching this play out:
AI doesn't fill the gaps in your org. It scales them.
Designers are saving ten hours a week. Content teams are producing 10x the output. But when you look at the actual business, there are no noticeable results.
MIT reviewed over 300 AI initiatives. 95 percent showed no measurable P&L impact. These models aren’t bad, but we’re using them for an operating system built for a different era.
We are using AI to run the same old process faster
Marketing teams were designed for a world where content was expensive and slow. Every asset cost real money, so we built layers of review, and we put the CMO's judgment at the top as the last line of defense against brand risk.
That made sense before.
But the tools changed, and the workflow didn't. We made production almost free and left the factory exactly as it was.
History rhymes here. The first television commercials were just radio ads read out to the camera. They went nowhere. TV only started to work when advertisers stopped repurposing radio scripts and built for the screen instead.
Most of us are still using AI to run more radio ads on TV.
But we could be doing a lot more. The job is not to govern AI usage. It’s to rebuild the operating system so the work gets better every time instead of just multiplying.
When production gets cheap, the bottleneck just moves
When production gets cheap, the bottleneck doesn’t disappear. It moves. It moves to the review, the approval, the judgment call.
You can generate fifty campaign variants before lunch and still wait three weeks for them to be approved.
Faster production into an unchanged system is not speed. It’s a backlog of material. Material that could be earning business instead of waiting for a review - or being shipped without it.
Most reorgs scaled production while cutting headcount. One person cannot hold the judgment that ten people held. So as volume went up, judgment capacity went down.
Judgment is the scarce resource now
When everyone can produce everything, good enough is worthless.
Taste, restraint, a clear point of view, that is what’s scarce now. Design the work so AI carries the volume and your people hold the calls that carry the brand.
What rebuilding actually looks like
Marketing is a growth team, so how we work with AI needs to be built for growth. This means building a system where judgment goes into the work once, and applies every time after that, so you and your team don’t get stuck reviewing the same type of asset on repeat.
Here’s the model my team actually works from and how we set it up in Claude. Regardless of the tools you’re using, the same principles apply.
How the pieces fit
Reading from the top:
Humans own judgment. Positioning, what good looks like, the calls that carry brand risk. None of this gets handed to a model.
The Project is the shared brain. Everything else works from it. This is the single source of truth that stops your team from solving the same problem in 200 private chat windows. When the Project gets richer, every skill and every agent gets better. That is the part that compounds.
A Skill is a piece of craft written down once that teaches the model how to do one thing the way your team does it: The writing style you use for a product launch, the shape of a strong opinion piece, the way you talk to one segment. You build it once, you encode the taste, and then everyone on the team, and every agent, applies it the same way. That is your brand code turned into something reusable and scalable.
An Agent is the automation. In Claude, you build these in Claude Code and connect them to your real systems through MCP, the layer that lets an agent actually connect to your ad platform, your CMS, your analytics. This is where the repetitive work lives. Bidding on PPC. Refreshing old blog posts for search. Reporting. The agent runs it end to end, and it borrows the relevant skill so the output stays on brand.
What starts as a judgment call becomes a documented skill. What becomes a skill eventually becomes an agent. The work travels down. The system gets smarter. The agents take the volume, and the humans make the calls.
This is not a question of headcount.
We are not removing the humans and praying AI fills the space they left. We are rebuilding the workflow so people spend their time on the work only people can do.
At this point, I’ve had to stop myself from jumping into reviews that no longer need me in them. Old habit.
The real question
The teams that win this won't be the ones with the biggest stack or the highest output volume. They'll be the ones where quality doesn't degrade at scale because it's built into the system, not policed at the end of it.
The real question is, are you running your marketing faster or rebuilding it?