AI as CEO
Objectives in, decisions out. I don’t assign it work — it decides what the work is, then reports back to me.
CASE STUDY · FLAGS
$1M exit in three years.
That’s the goal at Flags — and I didn’t set it. The AI CEO did, after analyzing what’s possible in the market.
Flags is ZeroCrew’s portfolio company, built from day one to run on a zero crew. The AI builds the apps, designs them, runs SEO and paid ads, answers users. I have things to do to make the goal happen, but my role is mainly investor: time and cash. One review a week.
0
employees. Nobody has ever been hired.
1
meeting a week. Me and the AI, reviewing.
3
consumer apps live in the App Store.
12+
countries running paid search.
13+
App Store locales shipped and maintained.
THE ORG CHART
One meeting a week. That is my whole calendar.
The review looks like any board meeting. Where are we, what moved, what is next. The difference is which side of the table holds the pen: the AI shows up with the state of the business already assembled, and it leaves with objectives rather than tasks.
Everything the agents do between one Monday and the next gets argued against the $1M goal — the number the AI chose, so the number it has to defend.
I kept two things and handed over the rest. Product judgment — the shape of the thing, whether it feels right to put in front of people — and capital, the decision to fund it at all. The building and the designing aren’t mine. Neither of the two things I kept is a workflow, and that is the whole point.
Investor / Product Architect
I put in the money and the shape of the product. One meeting a week. No inbox there, no standups, nobody to manage.
A zero crew, on purpose
Not a lean team. Not a small team. No team — none of the six functions below has a person attached to it.
THE FUNCTIONS
Six functions. I run none of them.
“Run by AI” is an easy thing to write on a slide, so here is the same claim as a list of jobs that would sit on a payroll anywhere else — each one live, in production, today.
App building & design
Native iOS apps, written and designed by AI. Three of them are in the App Store right now, and every release since has gone out the same way. I haven’t opened a design file or a code editor for any of them.
Product decisions & research
What gets built next isn’t a hunch I bring to the weekly. The agents run the research first — demand, competitors, what users keep asking for — and the roadmap comes out of that. I get to argue with it. I don’t get to write it.
Marketing — SEO and paid ads
Both of them AI. Paid search runs in more than twelve countries on agent playbooks instead of a media buyer. Spend goes out in pounds, revenue comes back in dollars, and the agents reconcile the two keyword by keyword — down to which search term paid for which trial.
App Store operations
More than thirteen store locales kept current: metadata, screenshots, submissions. The part of publishing that quietly eats a person’s week is a scheduled job here.
Reporting
Seven performance reports a week — spend, revenue, unit economics — written on schedule and waiting before I ask for them. They get written whether or not I read them.
Customer support
Users write in. Agents answer. The queue doesn’t wait for business hours, or for me to come back from holiday.
Tuesday, nine in the morning. I bring one question to the review and leave with three decisions I didn’t make. Somewhere in the night the ads rebalanced themselves and a locale I have never visited got new screenshots. I own the money and the judgment. It owns the week.
THE PRODUCTS
Three apps, in the hands of real users.
None of this is a prototype or an internal demo. They ship to the App Store, take payment, and get reviewed by strangers who have no idea a zero crew built them.
Flags: Schengen CalculatorDays counted, borders respected. For anyone living against the 90/180 rule.
Flags: Tax Residency TrackerWhere you have been, and what that makes you. Residency thresholds tracked as you move.
Flags: Countries Visited MapThe travelled world, drawn. The one people open for pleasure rather than paperwork.WHAT I TOOK FROM IT
Zero is one answer. It is not the answer for most companies.
I pushed Flags to the far edge on purpose. It’s my control experiment: build a company where the core team is empty, and find out how far the operating system carries it on its own. Product, acquisition, store operations, reporting, support — all held by agents, all of it visible from the outside in an App Store listing and an ads account.
Almost no founder wants that, and I don’t think most should. My other company sits at the opposite pole. Kubrio is a one-person company — separate from ZeroCrew, same founder — and there I’m the CEO, deep in a lot of the processes, with agents working as helpers instead of owners. Same operating system, one seat filled instead of none.
Your company sits somewhere between zero and one. That’s the whole job of the audit: find your number and say it role by role — who is genuinely core, what runs on agents, where AI just makes the rest faster.
My other company — one person, a lot of agents →WHAT THIS MEANS FOR YOU
I’m not selling you Flags. I map how it runs onto your company.
None of this needed a breakthrough. It needed objectives in writing, playbooks an agent can follow, data an agent can read, and a human who knows which decisions to keep. That stack travels. Two weeks is what it takes me to draw it for a company that already exists — with people, revenue, and habits.
The AI Automation Audit is $2,500, delivered in two weeks: your whole team interviewed by AI, then a map of who is core, what moves to agents, and what to build first. We start with a discovery call, and if the fit is wrong I’ll tell you on that call.
Twenty minutes on a call. The audit is $2,500 — done in two weeks.