Why we don't publish your posts automatically
Everything that goes out comes in as a draft and waits for someone to press send. That's not a promise in a proposal but a boundary in the build, and the question that remains is what happens on the day nobody's looking.
Lees dit in het Nederlands →$ publish --require-human
Why we don't automatically publish your posts
Because the system doesn't have a key to do it with. Everything that goes out comes in as a draft and waits for someone to press send. Not a promise, but a boundary in the build: the roles that do the work have read access and write access to your drafts, and no publish access. That key is yours, and nowhere else.
This is an uncomfortable opening for a piece that's supposed to set us apart: almost everyone says this by now.
We thought so too, until we measured it. And then it turned out the principle wasn't our position but the consensus of an entire category. That's an awkward outcome to write down and at the same time the only honest place to start.
Because the question that remains isn't whether a human decides. That question is answered. The question is what happens on the day nobody's looking.
01 / the definitionWhat exactly does draft-only mean?
Draft-only means a system does all the preparatory work but never makes the final decision. The post is written, the email is ready, the proposal is built. Exactly one action is missing: sending. That action stays with a human, every time, with no exception button.
The difference seems small. One button more or less. Yet it's the difference between a system that works for you and a system that speaks on your behalf without you knowing. In the first case you save the preparation and keep control. In the second case your name is attached to texts you've never seen.
02 / the measurementDoesn't every agency say this by now?
Yes. We measured it on 26 July 2026 by literally asking an answer engine why AI agencies don't publish automatically. The composite answer came out as: everything comes in as a draft, a human makes the publication decision. Eight agencies in the source list. We weren't among them.
That was more instructive than a citation would have been. The principle isn't the differentiator, it has become the category standard. Everyone says a human decides, the same way every contractor says he works tidily.
A principle everyone states says nothing anymore. What it's worth becomes clear from where it's built in.
Because there's a difference between an attitude and a boundary. An attitude is a sentence on a website. A boundary is something that stops the machine, even when there's a rush, even when it would have been more convenient, even when nobody would notice. The first costs an afternoon of copywriting. The second costs a design decision you can't undo afterwards without rebuilding it.
And that's exactly where the friction is in this category. Automatic publishing is on wish lists here: as functionality still to come, as something clients ask for and that gets delivered "later". Whoever has it as a wish on their roadmap hasn't built the boundary. They've postponed it.
03 / the buildSo how is that gate built?
With keys, not agreements. The roles that do the work run on access that can read and prepare drafts. Publishing requires a different key, and that one isn't in the machine. It sits with the human responsible for what goes out.
That's how it's built for us too. We're client number one of our own system, and that's not a marketing line but the reason we can be concrete about this:
The nice thing about this setup is that you can verify it without taking our word for it. Ask what rights the automation has. Ask whether publishing runs on the same key as producing. A provider who doesn't give a sharp answer to that hasn't built it this way.
This post itself came in that way. Written by a system, set down as a draft, read and approved by a human before you could read it. Had it gone differently, you wouldn't have seen this.
04 / the boundarySo what is the system allowed to do on its own?
Everything without judgement and without tone. An appointment confirmation, a reminder, an acknowledgement of receipt, a status change in your own records. Messages that contain nothing anyone could read as your opinion. Those go out on their own, and that's how it should be.
The boundary sits where the content says something about you. Content, sales, advice, a reply to a human. Everything where a reader thinks: this entrepreneur thought or wrote this themselves. A human should be in that loop, however good the system gets.
That boundary is deliberately drawn wider than strictly necessary. It would save time to also let the standard replies go out on their own. We don't, because the exception wrecks the pattern: as soon as one category exists that goes out without a human, the question is no longer whether something slips past the gate but when.
05 / the catchDoesn't approving cost the time you wanted to save?
No, and that's the sum most people skip. Writing a good post yourself takes an hour. Reviewing a well-prepared post takes a few minutes. The saving is in the preparation, not in the button. Whoever saves on the button is saving on the wrong part.
But there really is a catch, and it's not this. The catch is that approving is work that comes back every day. Not much work, but daily. Whoever postpones it builds up a queue and then experiences the system as pressure instead of help. A fixed moment each day solves that. No fixed moment means a pile.
Set against that is something you only get by looking. Every correction you make is the input that sharpens the system's grasp of your way of writing. If you adjust the same phrasing three times, that adjustment should become the default. Approving isn't, then, the leftover bit of work at the end. It's the point where your judgement enters the system.
06 / useableWhat do you ask every provider about this?
Four questions that expose the difference between attitude and build. They work just as well on us as on anyone else, and that's the point.
- Does publishing run on the same key as producing? One key for both means the boundary is an agreement, not a boundary.
- Is automatic publishing on your roadmap? A yes means the principle is negotiable as soon as a client pushes hard enough.
- Which messages already go out without me right now? If the answer is longer than confirmations and reminders, the boundary is drawn somewhere other than you thought.
- What happens if I don't approve for a week? A system that then sends anyway doesn't have the gate.
We also flag what came about with AI support, and we sort out the instructions for you and your team. Not because we know the legal outcome (that's work for a lawyer, not for us), but because for every piece of output you need to be able to show who signed off on it. At the bottom of this article is exactly that kind of line.
Does the system send or publish anything without my approval?+
Can I turn on automatic publishing later after all?+
How much time does approving cost per day?+
Don't I then become the bottleneck of my own system?+
What's the difference between ChatGPT and your own AI system?+
Want to know where in your business a gate like this should sit?
Book a call
We'll look at where your time is leaking away and tell you honestly whether we can do anything about it. Often the answer is: you can do this yourself. Then we say so, and you've spent thirty minutes on a clear answer.
This text was produced with AI support and checked and approved for publication by a human.