Fewer Makers, More Connectors

I have built marketing teams for most of my career. First marketing hire at a Series A data company, building the function to fifteen people. A digital marketing team inside a multibillion-dollar corporation. An internal agency for a North American enterprise operation. I now run a marketing function where machines carry the production and no one reports to me. So when a CEO asks me how AI changes marketing hiring, I am not reasoning from a keynote. I have staffed the function both ways.

The answer most people expect is a smaller version of the old team. That is wrong, and it is wrong in an interesting way. AI does not shrink a marketing organization toward some lights-out operation. It moves the center of gravity. You stop hiring people to produce artifacts, because production is what the machine took. You keep hiring, and in some seats hire more deliberately than before, for two things: ownership of the systems the machines run on, and judgment the machines cannot form. And the budget the collapse frees up should go to the one thing that gets scarcer as AI gets everywhere, which is actual human contact.

One rule underneath all of it: AI takes the making, never the owning.

What I inherited

The fullest team I ever built started with nothing. The company had raised its Series A, shown product-market fit, and stood up a sales team selling risk intelligence to some of the world’s largest banks and government agencies. I arrived as the first marketing hire, into a function that did not exist.

I did not start by hiring. I started by doing the work myself, long enough to learn the company and find out where the value was. That became my method for every seat that followed: play quarterback, jump into a function personally, run it until it proved there was enough value and enough work to justify a dedicated person, then hire for it. Nothing went on the org chart on theory.

The team that method produced, in order: a director of digital marketing and an events manager first, because the sales cycle was face to face and the fastest route to the buyers ran through small industry and government events. Then a VP of content over a team that grew to about ten analysts, regional domain experts, people who spoke the languages, knew the politics, and worked our data daily. They were the content engine: investigations into cartel-owned businesses, sanctioned shipping, forced labor in supply chains, plus the webinars that became one of our biggest growth channels. Then an email marketing manager when that channel proved itself. Then a dedicated designer, because design had become the bottleneck and borrowing capacity from the product team was not keeping up. Outside agencies handled the heaviest brand work, including a full design system when we rebranded.

That team took the company from Series A toward its later rounds, and I would defend every hire in it. Which is what makes it the right thing to rebuild as a thought experiment.

What I tried

The honest way to answer the hiring question is to take that exact function, the biggest and best team I built, and staff it again today, on the technology I now operate. Role by role. What survives, what changes shape, what goes.

The analysts survive almost untouched. This is the part that breaks the naive version of the AI story. That team was not large because of production volume; it was large because there is no such thing as a general analyst. A person covering a region has to know the language, the culture, the politics, and the data. That is domain judgment, and it is exactly what the machines cannot supply. What changes is the workflow inside each expert’s day. You no longer hire writers, you hire editors, and those analysts were senior enough that they already were. The machine does research support and first drafts, the human supplies the direction and catches what the model invents, then edits in conversation with it, and different models run the final checks so the checker is never the writer. I would probably run leaner per region, one expert where we sometimes had two or three. But the seats survive, because what made them valuable was never the typing.

The designer does not survive, and the reason is the same rule pointed the other way. Design at that company was pure production bottleneck: requests queued, briefs explained, iterations stretched across two busy people’s calendars. Today, with a detailed brand book and a library of approved examples loaded as context, the machine designs to spec, and the conversation that used to take days happens in an afternoon. But the cut is only available if that foundation exists, and someone senior still has to own and evolve it, because brands go stale and new products arrive. In a startup that owner can be the marketing leader with a strong brand system behind them. At enterprise scale it is a visual director responsible for consistent application of the brand, sitting above tools rather than inside production. Either way, nobody gets hired to modify a deck or produce a graphic ever again.

And then there is the role I once refused to hire at all, which is where this piece stops being a thought experiment and becomes a confession.

At that data company, my director of digital marketing kept telling me we needed a marketing operations person. I did not see it. The role was newer then, I had never worked alongside one, and I believed operators should own and maintain their own infrastructure. So I did not fight for the hire, and I know it frustrated him. He was ahead of his time, and I was wrong in the exact direction that has since become expensive.

Years later I made the same mistake in a different costume. When AI arrived, I was the excited one, and my most senior content person was the resistance, telling me flatly that this stuff produces slop and everyone can smell it. He was right about the failure mode. I underbuilt the foundation at first, pushed output without the context and instructions and examples that make the output worth reading, and got burned before I accepted what he was pointing at. Both mistakes are one mistake. Twice I treated infrastructure as overhead when it was the job itself, and twice someone on my own team saw it before I did.

What I would do now

Today the operations seat is not one I would merely accept. It might be the first hire I would fight for, and not for the traditional reasons alone. The clean data, the enrichment stack, the integrations, all of that still matters. But the real work now is bigger: connect the company’s data so that AI can query it, and the team stops reading dashboards of lagging indicators and starts asking the pipe questions in real time, with agents watching for problems and proposing answers a human reviews. Someone has to build and own that. The role I could not see the point of has become the one the whole operating model stands on.

Then comes the question nobody budgets time for: what do you do with what the collapse frees up? Fewer production seats, no deck-builders, no dashboard-readers, a leaner content bench. The default answer is savings. I think the default answer is wrong.

Spend it on human beings. As AI fills every channel, the internet fills with machine-made content, and every support line becomes an agent, the scarce resource is a real person showing up. So the roles I would add are the ones the efficiency conversation never mentions: an events person, because rooms where your users meet each other cannot be automated. A community manager, because I have built an online community and watched what it does for retention that no campaign matches. A social presence that behaves like a person in the places your customers actually talk, not a brand broadcasting into them. Partner and affiliate relationships, because when nobody trusts machine-made content, validation from an actual human carries a premium. That is where the freed budget goes: out of production, into connection.

So the marketing organization I would build now barbells. Machines in the middle, carrying production at volume. On one end, a small number of people who own systems and judgment: the ops person running the infrastructure, the domain experts editing what the machine drafts, a senior owner of the brand foundation. On the other end, people whose entire job is being human with other humans, at events, in communities, in conversations no model should touch. Fewer makers, more owners, more connectors.

The rule that generates all of it is the one I keep returning to: AI takes the making, never the owning. Every hiring mistake I have made in this territory, refusing the ops person, underbuilding the content foundation, came from undervaluing ownership of the system relative to production of the artifact. The machine has now made that error impossible to afford. Staff the owning, automate the making, and spend what you save on the people whose job is other people.