PA PubAgent Limited Hospitality Consultancy

Notes from the Frontier · AI strategy · operations

I told a client how big my team is. They went quiet.

A throwaway line about who actually does the work stopped a client cold. What running a business with a genuine round-the-clock AI workforce looks like from the inside, and what it should make you ask your own suppliers.

Joe Mullane, Director of Divergent Thinking · · about 2 minutes to read


A cluster of small navy circles orbits one larger orange circle at the centre

A client asked, almost in passing, who does the work behind what we deliver. I told them. The room went quiet in the way rooms go quiet when a number lands that people weren’t expecting.

I don’t run a large office. I run a small number of people and an agent workforce which, together with its skills, runs into the hundreds. Each role has a defined job, an owner and a review point, and several genuinely run around the clock rather than nine-to-five. That’s the number that surprised the client, not because it’s a trick, but because most businesses talking about “using AI” mean a chatbot bolted onto one task. This is closer to a real org chart, just not a human one.

A role, not a tool

The difference that actually matters isn’t the headcount. It’s that each of these roles has a job description, the same way a person’s would: what it owns, what it’s allowed to decide on its own, and what it has to hand up. One of them gets a formal review, on a written record, the same as a person would get an appraisal. That structure is what makes “AI does this” a true sentence instead of a marketing one.

Round the clock doesn’t mean unsupervised

The part that should reassure a client, not alarm them, is that continuous operation sits inside the same review structure as everything else. A role running overnight still has an owner, still gets checked, still has a place its output lands where a human looks at it. The clock doesn’t stop. The accountability doesn’t either.

What leaders should do next

  • If a supplier tells you “AI does this,” ask them to show the structure behind it: who owns each part, what gets reviewed, what happens when something fails. A real answer sounds specific. A slogan doesn’t.
  • Apply the same discipline before you scale AI inside your own business: name the role, name the human owner, set a genuine review point, before you set it loose on real work.
  • Don’t be impressed or alarmed by a headcount number on its own. Ask what each part of it actually does for the outcome you’re paying for.
  • Round-the-clock is a claim, not a comfort, until you know what “review” means for a role that never clocks off. Ask that question of anyone who tells you it runs while they sleep.