For most of my career I ran digital marketing teams the traditional way. At Marsh McLennan that meant 14 people across time zones, a multi-million budget, and a governance model that decided who could publish what, when, and against which standards and approval structures. When you support 250 marketers across 12 operating companies, governance isn’t bureaucracy. It’s the only thing standing between a global brand and chaos.
Then, in 2026, I became a team of one.
I’d taken redundancy after a restructure and set myself a test: could a single person, using AI, run the operating model of a proper marketing organisation? Not “use ChatGPT to write posts faster”. Actually run the functions – business intelligence, content production, quality control – as an organised system.
Eight months in, here’s the honest answer: yes, but not the way most people are doing it.
The org chart
My company has an org chart. I’m the only human on it.
There’s an insights function. Twice a week, an agent queries production data and delivers a report to my project workspace – a deep read on Mondays, a pulse check on Thursdays. I didn’t build this because it’s clever. I built it because at enterprise level I learned that the teams who review performance on a rhythm beat the teams who review it when they remember to.
There’s a content function. Once a week, an agent produces article drafts against a queue of researched search queries. It never publishes. Not because the technology can’t – because I decided it won’t. Every draft goes through review before anything reaches the site, the same way nothing went live across 700 domains at a Global risk management corporation without sign-off.
And there’s a quality gate, which is the part people find strange and I consider the most important hire I made. It’s an agent whose only job is to check every piece of output against a canonical source of truth – approved claims, approved wording, a list of banned words and patterns that make writing smell like AI. Output that fails the gate never reaches me. My review time gets spent on judgement, not proofreading.
Delegation with governance, not delegation with hope
That’s the transferable principle, and it’s twenty years old. It just used to apply to people.
When you hand work to a new team member, you don’t say “do marketing” and walk away. You define the scope, the standards, the checkpoints, and the things they must never do without asking. Then you review the output until trust is earned. Nobody calls that micromanagement. It’s management.
Most solo founders I see using AI skip all of that. They delegate with hope. The agent gets a vague brief, full autonomy, and no quality standard – and then they’re surprised when the output is generic mush or, worse, confidently wrong. The problem isn’t the model. It’s that nobody would run a human team that way and expect it to work either.
The counterintuitive part: the smaller the company, the more the governance matters. At Marsh, a bad draft got caught by multiple layers of review before it embarrassed anyone. Here, there’s one layer, and it’s me. The canon file, the quality gate, the no-auto-publish rule, that’s me rebuilding the safety net at a scale of one.
What it costs, what it returns
The whole system takes me under an hour a day to operate. Review the insights report over my mid morning coffee. Edit and approve drafts in a batch. Make the calls only a human should make. The preparation, the shortlists, the research, the first drafts, the reporting – happens while I’m doing something else…or out walking the dog.
Is it the same as a 14-person team? No. There are things a good human team does that no agent replaces, and the biggest is telling you you’re wrong before you’ve asked.
But it produces the output of a small department at the cost of one determined person, and it does something else I didn’t expect: it made me a better operator. When every process has to be defined clearly enough for an agent to execute it, you find out fast which of your “processes” were actually just habits.
Twenty years of enterprise marketing taught me how organisations should run. Building one alone taught me which parts of that knowledge were load-bearing. Turns out governance was never the paperwork. It was the product.
I write about digital operations, transformation, and what happens when enterprise discipline meets a team of one. If any of it is useful to your organisation – or you’re hiring for someone who thinks this way – the contact form is on the About page.