The job
“Our customers know they should be publishing. None of them will. Can we sell them the cadence instead of the advice?”
Content is the offer everyone wants to resell and nobody wants to staff. The work is not the writing — it is the weekly decision about what to write, the research behind it, the image, the distribution copy, and the person who has to care on a Friday afternoon. That is why content retainers are sold, badly delivered, and quietly cancelled.
What the run actually does
The subject is chosen from what the researcher found, not from a content calendar somebody has to maintain. The draft arrives with its hero image and its distribution copy attached. And it arrives as a pull request — the single design decision that makes this sellable, because a piece of automated writing under your customer’s brand should not be able to reach the public without someone saying yes.
The proof is the site you are reading
This argument would be weak from a vendor that did not run it on itself. Most of the posts on this blog were committed by a bot on exactly this workflow, and every one of them was merged by a person. You can read the commit history rather than take our word for it.
Why it needs nothing from us
Site generation needed a platform change before a partner could adopt it. This does not. A publishing cadence is a workflow, two characters and a linked repository, inside your organisation — all of which you can create yourself today. Two of the characters you would want (a marketing lead and a web researcher) already ship as presets to copy.
That makes this the fastest of the three to stand up for a customer, and the easiest to shape per customer: a different voice, a different subject area, a different repository, same machinery.
Honest state
Running in production, weekly, on our own site and on customers’ repositories. The full path — schedule, research, draft, image, distribution copy, pull request — is shipped and has been running for months.
The limits worth setting expectations on: the repository has to be one the workflow can be linked to, so this fits customers who have a git-backed site and not customers who are on a hosted page builder with no repository behind it. A merge still needs a human, by design — if your offer implies untouched-by-hands publishing, this is not that, and we would rather lose the deal than have you discover it in month two. And a voice is only as good as the character you write: the preset is a starting point, not a substitute for telling it what your customer sounds like.