I let my AI post to my channels without me sitting there. But it only goes when I say go, it never emails without permission, it checks its own work before it publishes, and it stops cold the moment a channel is paused. Here is how the guardrails actually work.
Every content creator hits the same wall.
You build a system. You get it running. Then you realize someone still has to sit there and push publish. Every time. On every platform. While also writing the next thing.
That was me six months ago.
I run 33 brands. Each one has a posting schedule. Some post to Substack. Some post to X. Some post to Instagram. Some post to all three on the same day.
I was the one checking every post before it went live. I was the one flipping between tabs to confirm the right draft was loaded. I was the one double-checking that the Substack post did not have the "send via email" box checked, because my list does not want an email every time I publish a note.
It was not hard work. It was just constant, repetitive, low-brain work.
The kind of work that eats your afternoon without you noticing.
I did not want to stop reviewing content. I wanted to stop doing the mechanical stuff around the review. The channel checks. The duplicate checks. The email toggle. The "is this post stale" check. The proof step before anything goes live.
So I handed that whole layer to my AI.
I built a dispatch system inside ATHENA, my content brain. It runs the mechanical checks. It holds a lease on each post so nothing runs twice. It reads channel state before it touches anything. It requires a proof pass before the Bitchwork articles publish. And it will not send a Substack email, ever, under any condition, because that rule is hard-coded.
But here is what I want to be clear about.
My AI does not post without my approval. It cannot. The approval gate is the first check in the whole chain. If I have not said yes to a piece of content, the dispatch stops before it even looks at the channel.
What changed is everything that happens after I say yes. That part runs without me.
What I can prove
These are not claims. These are the actual guardrails in my live system.
Approval state is the first gate. Nothing dispatches unless the content record shows it is approved. The system reads that before it does anything else.
Channel state is the second gate. Each channel has a live or paused flag in channels.json. If a channel is paused, the dispatch skips it and logs why. I do not have to remember which channels are live. The system knows.
Substack send:false is hard-required. The posting script for Substack has one rule it cannot bend: send_email must be false. It will not publish a Substack article if that flag is missing or wrong. I do not have to remember to uncheck the email box. The code refuses to run without the right flag.
A proof step is required before Bitchwork publishes. Before a B*tchwork article goes live, the system generates a staged preview and checks it. That is the proof pass. Live posting does not happen until the proof passes.
The dispatch lease prevents doubles. The state file holds a lock on each post while it is in flight. If something runs, crashes, and tries to run again, the lease stops the second attempt. The same post cannot go twice.
Stale sources get refused. If the content was approved but the source data is too old, the system flags it as stale and does not post. It does not guess. It stops and waits for me.
All of that happens automatically. After I approve.
Behind this paywall
Paid readers get:
The exact approval handoff I use before any post goes live.
The step-by-step of how the dispatch chain runs after I say yes.
The real channel state, lease, and proof checks with plain-English explanations of what each one does.
The parts my AI still cannot do without me, listed plainly.
What to check manually every time, even with automation running.
A copyable setup checklist if you want to build the same guardrails into your own system.
What paid subscribers get every week
This is not the only B*tchwork article you get.
Every week you get:
Real work I handed to my AI inside my actual business.
The prompts I used, written so you can copy them.
The receipts that prove what worked.
The problems that slowed the job down.
The safety checks that kept AI from making a risky call alone.
Simple playbooks you can run in your own business.
Subscribe now and the rest of this article opens right away.
The full walkthrough is below the paywall.
Click the button, subscribe, and the locked section below opens right away. You get this article now.
