I recently read about a company hiring a chief marketing officer with no direct reports whose job would be to manage 3,000 AI agents.
I said it would fail miserably.
I was asked why I was so sure, which is fair, because I run a marketing organization that has gone about as far into AI as any I’ve seen. Ninety-three percent of our marketing employees have coded and built their own agents and use active AI workflows in their daily work. We’ve built and put into production over 20 distinct AI systems in 12 months: agents, automations, infrastructure and even internal replacements for software we used to license.
So, was I a bit of a skeptic when I heard about that job? Absolutely. Call me crazy, but I just don’t see the day when a CMO simply and successfully manages 3,000 AI agents without proper support and staff.
Good marketing and good human relationships are more complex than that.
Output was never the bottleneck
For as long as I’ve worked in marketing, we’ve operated under a scarcity of production. Not enough hands to write the emails, build the pages, cut the video and pull the report. Every process, budget and org chart in this discipline is a response to that scarcity.
AI has largely removed it. A much smaller team can now produce what a much larger team used to. And because our instincts were formed during scarcity, the reflex is to celebrate the volume and then quietly expect more of it. More content, more variants, more channels. The AI-powered marketing team is expected to do more. And it can, so it does.
That exposes something the industry keeps not saying out loud: nobody automated the decision about what to make. We now have effectively unlimited production capacity pointed at a strategy still scaling at human speed, and, in a lot of organizations, that strategy function is now the least-resourced part of the operation.
That’s because everyone got redeployed to feed the output.
Seems like that is a recipe for disaster, like a factory without a product manager.
The best decisions on my team last year was a judgment call no agent could make
My creative director put it more clearly than I could. Using AI, his team gets to roughly 85% to 90% of what they’d consider quality work. But it takes an hour and a half instead of two days. His conclusion: the hour and a half is what makes sense.
Read that again, because it’s the most important sentence in our entire AI transformation. He is deliberately shipping at 85% because in a market where relevance decays daily, speed beats polish. A LinkedIn post that goes up at 11 a.m. can have a video response live that same afternoon. That was not possible for us a year ago.
Now, which part of that could an agent have decided?
Not the tradeoff. Not the threshold. Not the read on which projects deserve the extra day and which don’t. That judgment requires knowing the market, the brand, the audience and what this particular moment is worth and not evaluated by a model that has ever seen your business.
His own self-reported AI comfort level, by the way, is a 6 out of 10. And, that’s simply because every time he digs deeper into AI, he realizes how much more he has to learn. That’s what active engagement looks like, and it’s worth more than a 10 from someone who stopped being curious.
What 3,000 agents actually costs you
Set strategy aside and look at the operations.
Every agent encodes a set of decisions about your business, made by someone, at some point, based on some understanding of context. Multiply that by thousands and nobody can answer basic questions. What is this agent doing? What data does it see? How many tokens are we using? Why did it recommend that? Who approved the version currently running?
A 10-person team you can’t fully see into is 10 people. Ten people each running hundreds of AI agents is thousands of processes nobody can see into. That sounds like a scaling and coordination nightmare that has to be solved.
And in any company with history, the 3,000-agent model destroys the thing that makes marketing actually work: domain knowledge. The nuances. The reason we don’t run that offer in that segment anymore. The customer story that can’t be used in that vertical. None of that is in your data. It’s in the heads of people who’ve been paying attention.
Then there’s accountability. If an agent creates a campaign or landing page that damages the brand, someone will answer for it. “The model did it” will not survive contact with your board. In the 3,000-agent model, the CMO is responsible with no visibility. In addition, most mistakes are already being made by the time the CMO is aware.
What the CMO role actually becomes
AI makes the CMO role harder, not smaller, and shifts the demands toward areas that current marketing dashboards don’t measure.
Brand consistency has to become infrastructure. When your team and dozens of systems are all generating in your voice, a style guide in a shared drive is not a control system. We built a knowledge graph for messaging, product, content, brand and design. That’s how the same message comes out of the organization regardless of who or what generated it. Most companies scaling AI output have no equivalent, and they will find out the expensive way.
Specialization survives. This surprised me. I don’t need everyone on my team to become an AI engineer. I want my events person to be excellent at events and my creative director to be excellent at creative. What they needed was someone to sit with them and help them learn how to excel using AI in their specific vertical.
Someone has to own AI as a full-time job. Not a third of somebody’s role. We embedded a dedicated AI lead in marketing for a year … and it worked. The clearest evidence it worked is that he’s since moved to a company-wide role and the marketing team runs all of it without him. I don’t think having dedicated AI lead is necessary for the long-term. But in the short term, it’s the only way we could get our entire marketing team to actually use AI effectively, solving challenges they individually faced.
The constraint is not technical talent. Our best automation story is a non-coder in events who reclaimed 18 hours a week. If that’s possible, anyone can excel with AI – as long as the investment is made to enable them in the right way.
The marketing model that works
At our biggest event this year, we produced the keynote visuals internally and avoided over $50,000 in agency costs on that work alone. We also opened the show with a completely AI-generated video that had people in the audience in tears. All of the footage, effects, audio, etc. was made by AI, but the creative design was made by humans who understood story, emotion and what that moment needed. The theme of that film was that the most powerful force was never technology. It was human.
AI handles volume and pattern-finding at a scale no human can match. Humans supply judgment, taste, context and accountability.
Remove either side and the output degrades in ways your metrics won’t catch until a customer does.
So am I wrong? Is the future of marketing leadership in one marketing leader and 3,000 agents?
I seriously doubt it. Is it smaller teams? Potentially.
What the future definitely is, are valuable marketers whose impact is multiplied by the power of AI. You need the people. You need the systems. You need the AI. You need them working harmoniously together.
AI is not going to take every job. But it will take the jobs of those who don’t effectively figure out (and quickly) how to use it to multiply your impact.
Opinions expressed by SmartBrief contributors are their own.
____________________________________
If you liked this article, sign up for the SmartBrief Marketing Leadership Lab, a twice-weekly newsletter that distills market dynamics into C-suite action.
The post How AI is changing the CMO role appeared first on SmartBrief.
