The First Thing a CMO Must Do to Get a Marketing Team Actually Using AI Isn't Buying Accounts

A while back, I had dinner with a friend who runs marketing.

He looked deeply worried.

His company had spent good money buying the marketing team a pile of AI accounts — large language models, image generation, productivity tools, the works. Three months later, he logged into the admin dashboard to check, and the people actually using them? You could count them on one hand.

He asked me: "We spent the money. We held the meetings. Why won't this thing move?"

After hearing him out, I told him something he didn't much want to hear:

You think the problem is the tools. The real problem is the people.

The Hard Part Isn't Picking a Model — It's Rewiring Minds

What does "AI transformation" even mean?

For a lot of people, the picture that instantly pops into their head is: buy accounts, run training, roll out a system.

Wrong.

Those are surface symptoms. Another CMO I know runs the marketing team that uses AI most fluently, and she told me something I still remember —

"The hardest part is never which model to pick or how many accounts to buy. It's changing how the team thinks, how they write briefs, how they review work, how they hand things off."

Think about it — isn't that true?

Tools don't move on their own; people do. If somebody genuinely believes deep down that "this stuff isn't reliable, I'd better do it myself," you could hand them the strongest model on the planet and they'd open it twice and then set it aside.

So the CMO's number-one job isn't to draw up a training plan, isn't to build a KPI spreadsheet, isn't to bring in a consultant.

It's to make the team want to use it first.

Making the Team "Want to" Starts with One Low-Pressure Sharing Slot

So how do you make people want to use it?

I'll give you a remarkably plain method.

In your regular weekly team meeting, carve out ten minutes and ask everyone one question:

"This week, did AI save you any hassle? Even once."

Don't turn it into a "prompt-sharing conference." That stuff is too dry — dry enough to put people to sleep.

What you want to ask about is concrete, human stories —

A colleague found a research path that used to take two days; with AI's help it got done in two hours. A certain agent that's almost usable — what's the gap, let's puzzle it out together. Even a failed attempt, as long as somebody genuinely tried, is worth bringing to the room.

Praise these things openly and generously.

At the same time, pull out the little nails quietly killing AI enthusiasm, one by one.

The most common nail: account quotas. The quotas are meted out so tightly that everybody has to do a mental calculation before each call — "is this call worth it?" Think about it: in that state, who's going to feel free to experiment?

Don't make the team count tokens every time they want to try an idea.

Once three months pass and a few genuinely curious "seed players" emerge in the department, with a handful of real, won projects under their belts — that's when you take the best few cases and codify them into the department's standard workflows and automation templates.

At that point you're no longer "forcing" something new. You're taking what already works and formally writing it into the rules.

It feels completely different.

Training and Hands-On Practice Must Start the Same Week

At this point somebody might say: but surely we need training, right?

Right, you need training. But I've seen too many companies where the training is grand and ceremonial, the staff finish it, go back to their desks, and keep working exactly the way they always did.

Why?

Because if knowledge doesn't land as a concrete deliverable, most of it evaporates within a week.

That CMO who uses AI best set one ironclad rule —

Every training session must be paired with a real project that kicks off that same week.

What does that mean? You can't come back empty-handed from the training. You have to immediately grab a piece of work on your plate and try it. If it works, it's yours. If it doesn't, you explain why at next week's meeting.

Just listening to lectures without getting your hands dirty doesn't count.

Before the Process Is Smoothed Out, AI Only Amplifies the Chaos

Having covered people and training, the next point is one I especially want to emphasize.

A lot of people think: our processes are messy? Bring in AI — let it sort us out.

Wrong. Dead wrong.

If a process is already messy, AI won't tidy it up for you. It will only take your chaos, amplify it, accelerate it, and replicate it at scale.

The right order is the reverse — first smooth out the process, then let AI step in.

How do you know which processes are worth smoothing first? Don't go by gut.

Look at where the time goes. Use time-tracking data to see where the team's hours actually landed this week. The tasks that are high-frequency, repetitive, and rule-based — those are exactly where AI most deserves to take the field.

Every quarter, put together a "processes-to-optimize list." For each item, spell out clearly: what does it look like today, what should it look like, who owns it.

Once this is done, the focus of the team's discussions shifts.

It's no longer "where can AI be used" — that aimless, pie-in-the-sky debate.

It becomes "given this bottleneck, how does AI solve it."

That is the position a CMO should be steering the team to.

Top-Down Push and Bottom-Up Pressure Have to Happen at the Same Time

Direction alone isn't enough. When you actually try to land it, you'll find one thing —

Pushing from a single direction doesn't move it.

You need both directions at once.

Top-down: the leader has to be a visible practitioner themselves. Not just mouthing "everyone should be using this," but dropping the prompts they've written, the agents they've built, the drafts they've red-lined straight into the team chat. If you don't use it, why should the people below you?

Bottom-up: identify the people on the team who are genuinely curious, the ones noodling on this in their spare time. Hand them the stage — let them present in meetings, praise them publicly. They are the sparks. Once the sparks catch, the blaze follows naturally.

My CMO friend has a goal I find genuinely elegant —

By year-end, have those seed players take the best cases running in the department and turn them into a "Marketing AI Field Manual."

What should be in it? Prompts, processes, QA rules, agent logic, data conventions, and the pitfalls they hit.

This manual is worth more than any external course.

Because it grew out of your company's real work.

Hiring a Consultant: Don't Pay for a Title

Should you bring in a consultant somewhere along the way?

Most of the time, yes. But who you hire makes all the difference.

What you need isn't somebody who's "worked on an agent" — people like that are everywhere now.

What you need is somebody who has solved a problem like yours.

What does that mean? This person has to genuinely understand the marketing tech stack, understand marketing workflows, understand how a marketing team actually runs day to day. They have to be able to help you straighten out the workflow, wire the tools together, and pull the team from flailing between one experiment and the next into a state where they can execute, repeatedly.

Don't pay for a title — pay for relevant experience.

Don't Fool Yourself with Vanity Metrics

Finally, let's talk about the step most likely to derail — measurement.

Using the wrong KPI is worse than having no KPI at all.

What's a vanity metric?

How many prompts were run, how many training hours were clocked — these numbers look pretty but don't tell you a single thing.

For a KPI to work, it has to tie back to the business and to actual work output.

Let me show you the scoreboard my CMO friend uses, so you can feel what "forcing the team to do the right things" really looks like —

• The pilot team: every member completes the required AI courses.
• Each person, after training, delivers one real AI-assisted project.
• All high-frequency processes are mapped out clean and prioritized.
• The consultant runs at least two hands-on build sessions.
• At least three agents are running in production.
• The top three daily tasks have their time-to-complete cut by 60%.
• The pilot team's weekly active usage rate tops 70%.
• The reporting delivery cycle is cut in half.
• Content output grows 5x without adding headcount.
• 80% of what AI produces only needs light edits to be usable.
• A reusable field manual is captured.
• In the year-end retrospective, AI's contribution gets factored into ROI.

See the pattern?

This set of metrics doesn't reward "how much was learned" — it only rewards "how much labor was saved, how much useful output was produced, and how the results were proven with data."

KPIs are the conductor's baton for behavior. Point it the wrong way, and the whole orchestra plays out of tune.

So What Should a CMO Actually Do?

Having said all this, I really just want to make one point.

Why does AI transformation fail?

Because it so often gets slotted in at number eleven on a long priority list — the ten things ahead of it are all more important, and it never gets its turn.

It only succeeds under one condition: the CMO treats it as a serious deliverable, gives it time, gives it budget, gives it training, and treats "the first three agents that actually run" as project milestones — not as a hobby to dabble in on the side.

Spark curiosity, build confidence through hands-on practice, clean up the processes, lift up a batch of sparks, and nail the KPIs down.

Walk that path, and what you end up leading isn't a marketing team that's merely "using" AI —

It's a marketing team that runs on top of AI.

May you build a team like that before long.