One marketer, five properties, 100,000 guests

I run a hospitality group. Five properties, glamping sites and hotels, somewhere north of 100,000 guests a year. The marketing team is one person.

That number gets quoted back to me in a way I want to correct, because the interesting part is not the headcount. It is which decisions stayed with the human.

What the agents actually do

Three workloads, and they are not equally hard.

Content production across five properties. Each property has its own character, its own season, its own reason someone books it. Producing enough material for five of those, continuously, is what used to consume a team. This is the workload agents handle best, because the constraint was never judgement. It was volume.

The influencer programme. Over 1,200 collaborations a year. Outreach, scheduling, briefing, follow-up, the long tail of people who said yes and then went quiet. This is the piece people are most surprised by, and it is the piece I would defend most strongly, because 1,200 relationships a year is not a creative problem. It is a coordination problem with a thousand small deadlines, and a thousand small deadlines is exactly the shape of work a human is worst at holding.

Guest communications, at that guest volume. Questions before booking, questions during a stay, the same fifteen questions in a hundred phrasings.

What the human kept

Everything upstream of those three.

She decides what each property is *for* this season — who it is aimed at, what the offer is, which of the five gets pushed when. She decides which influencer partnerships are worth having, and that is a judgement about brand fit that has no reliable proxy. She reads what goes out. She is the one who notices when the tone has drifted.

The line, roughly: agents own execution and coordination, the human owns intent and taste. Where it gets blurry is where I pay most attention, because that blur is where the failures come from.

The part that does not transfer

I have watched people take this number and reach a conclusion it does not support, so let me draw the boundary myself.

Hospitality has properties that a lot of businesses do not:

• The questions repeat. A guest asking about check-in time is asking something with a correct answer, and the answer is stable.
• The content is high-volume and moderate-stakes. A post that lands badly is a bad afternoon, not a compliance event.
• Influencer coordination is genuinely mechanical once the partner list is decided.

Change any of those and the maths changes. If your customer questions each require reading a contract, this does not transfer. If a wrong sentence is a regulatory problem, the review step you need costs more than the production step you saved. The businesses where this works are the ones with a lot of repeated, moderate-stakes, coordination-heavy work — and that is a real category, but it is not every business.

What I got wrong first

The early version tried to hand over judgement along with the work. Let the system decide which property to promote, which partner to approach, what the offer should be. It produced output that was defensible in every individual case and directionless in aggregate.

The fix was unglamorous: pull the decisions back to the human, keep the execution with the agents, and be explicit about which is which. That boundary is now the thing I check when a workflow starts producing noise. Nine times out of ten a decision has quietly migrated to the wrong side of it.

Why I am writing this down

Because "one marketer, 100,000 guests" is the kind of line that sounds like a product claim, and I would rather publish the shape of it than let it stand as a number.

The honest version is narrower and more useful. One person can direct an operation this size when the work underneath is high-volume, repetitive and coordination-heavy, and when that person still owns every decision about intent. Take away the second half and you get five properties producing a great deal of content that means nothing in particular.

We build this kind of thing for other companies now — that is what forward-deployed engineering is, in our case: engineers who work inside your team and ship into your stack rather than handing you a strategy deck. If you want to know whether your operation has the shape described above, that is roughly what the first call is for.