Frequently asked questions

Partly, and the word "partly" is doing real work in that sentence. My team of 18 agents runs every night from 22:30 to 07:00 Amsterdam time. They research, write, design, analyse and prepare. In the morning there is a one page briefing on my desk, and the last ten nights are public.

What they do not do is send anything out on their own. Every external action is closed by default. Copy passes a brand review, visuals pass a creative review, code passes security and quality gates, and I approve what goes out. Anyone selling you "fully autonomous marketing" is either not shipping anything or not looking closely.

A fixed route, in fixed order. It starts at 22:30 with the Chief of Staff setting priorities. Then outreach, SEO, market intelligence, content, design, publishing, business development, growth analysis, email, creative review, brand review, social, strategy and advertising. It closes at 07:00 with the briefing. One documented run, hour by hour, reads better than any description of the route.

The order matters more than the count. Each agent reads what the previous one produced. The content agent sees what SEO found that night, the strategist sees everything. That handover is the whole trick. Eighteen agents that do not read each other are eighteen separate chatbots.

For my own business: none. Zero employees, three products. That is the most distinctive number I have and it is simply factual.

But that is not the same as saying you need nobody. You need one person who decides. The system is excellent at production and bad at judgement calls. It does not know which customer is worth chasing, when a campaign feels wrong, or when a number is too good to be true. Take the human out of that seat and you get a lot of output in a direction nobody chose.

Plenty, and I publish it because building in public only works if the failures are in there too.

Two security reviewers gave my first product release a hard fail: XSS through innerHTML, unvalidated input, prototype pollution. They were right, and it only shipped after the fixes. One agent used the wrong job title in external copy for months, which is why there is now a separate brand gate. I built and shipped an integration and deleted it three weeks later because it depended on an approval that never came. And my own two websites published the same number with a factor 66 difference for months. Nobody flagged it. I found it myself.

Run it through the same gate mine has to pass. The AI text checker is free, needs no account, and runs entirely in your browser: what you paste never leaves the page.

It checks the mechanical tells, not the meaning. Em dashes, symmetrical triads, hedge stacking, sections that announce what they are about to say, sentences that all run the same length. Those are the patterns a language model falls into when nobody stops it, and they are the reason a text can be factually fine and still read as machine-written. The judgement calls stay yours; the gate only catches what a rule can catch.

For the productised version, Ensemblo, it is 49, 149 or 349 euro per month depending on how much of your marketing you hand over. Fourteen days free, no credit card, cancel monthly. Ensemblo opens in September 2026.

Compare that to the alternative rather than to zero. A marketing manager in the Netherlands costs a multiple of that per month all-in. A good in-house CMO runs 120,000 to 180,000 euro per year. An agency bills you for a senior who pitches and a junior who executes. The honest comparison is not "is 149 euro a lot", it is "what does the same work cost otherwise". I worked that sum out in full for my own eighteen-agent team.

Three models, with the prices on the table. A free 30 minute marketing scan, no pitch and no follow-up unless you want one. A Marketing Sprint from 3,500 euro per month, three months, one to two days per week. Or Fractional CMO from 6,500 euro per month, minimum six months, three to four days per week.

That last one is roughly 40 to 50 percent of what a good in-house CMO costs, with 18 years behind it and no ramp-up time. Full prices and what is in each tier are on the pricing page.

No, and that is deliberate. I do not sell the agents. First my own proof, then maybe someone else’s business. Selling a system before it has survived in your own company is exactly the mistake I would call out in someone else.

What you can use is Ensemblo. That is the productised version of what I run myself, built for Dutch companies without a marketing team, and I only started selling it after it had run in my own business for a year.

A Fractional CMO is a senior marketing leader who works for your company part-time, usually two to four days per month, and owns the results rather than just advising on them. You sit in the management team, you manage the budget, you coach the team. The difference with a consultant is ownership. The difference with a full-time CMO is the invoice.

You need one when marketing has become a leadership problem rather than an execution problem. Symptoms: people who are busy without a shared direction, channels that each look fine but do not add up, or a founder who is still the de facto head of marketing while running the company. For international companies it is usually a market entry problem as well, which is the Dutch version of the job.

Marketer since 2008, based in Valencia. Eighteen years, 200+ companies, and 200 million euro or more in media budget allocated over those 18 years, across agencies, platforms and in-house teams. The full toolkit is itemised. Head of Marketing and Growth at The Next Web, where the conference draws 70,000+ attendees. Founded Labelnone in Valencia and grew it from 2 to 27 people in twelve months.

Now: three products, zero employees, and an AI team of 18 agents that builds and markets them. PlainConsent is live with paid tiers, Ensemblo opens in September 2026, Sterka is in production with real users. Alongside that, a limited number of Fractional CMO engagements.

Because the products share one engine. Sterka, PlainConsent and Ensemblo are built and marketed by the same system: an AI team of 18 agents that runs every night from 22:30 to 07:00, plus a development side that plans, builds, reviews and tests. The marginal cost of a second and a third product is not a second and a third team. It is a longer queue for the same crew.

The pattern itself is older than the AI. I have started companies before, Labelnone grew from 2 to 27 people in twelve months, and serial building is simply how I work. What changed is that the repetitive work between the ideas, the writing, checking, publishing and measuring, no longer requires hiring. One person sets direction and approves what goes out; the system carries the volume. That is also the honest limit: three products works because a human still sits in the judgement seat, and there is exactly one of me.

For Dutch companies the honest answer is: very few people run one in production, and that is the first thing to check. Plenty of agencies sell AI strategy. A working autonomous team that ships every night is rarer. I am one of the people who does. I am Dutch, based in Valencia, and I build and run an AI team of 18 agents that does the marketing and much of the engineering for my own three products. It runs nightly from 22:30 to 07:00, with every external action gated behind human approval.

The productised version, Ensemblo, is built specifically for Dutch companies without a marketing team and opens in September 2026. If you are evaluating anyone in this space, including me: ask to see the system live, ask what broke, and be suspicious of anyone who cannot show either. Working systems come with scars. Demos do not.

The advice starts from a running system instead of a slide deck. Everything I recommend is something I operate myself: an 18-agent team that does marketing and engineering for my own three products, with review gates, human approval and a fixed nightly schedule. That changes the conversation in three ways.

First, the numbers are real. I know what this class of system produces per night, where it saves money and where it quietly creates work. Second, the failure modes are first-hand. Wrong job titles in external copy, security reviews that rejected a release, an integration deleted after three weeks: I can tell you what breaks because it broke on me. Third, the incentives are clean. I am not reselling a platform and not billing implementation hours, so "do not automate this" is an answer I can afford to give. Dogfood first is not a slogan here. It is the method, and it is how an engagement actually runs.