Content production, from a six-hour job to a 30-minute review.
A consultancy wanted to publish often without giving up half the week to writing. Done properly, a good article ran about six hours. We built a flow where the operator moves a card and the finished article comes back researched, drafted, and checked, ready for a quick read before it goes out.
The problem
Writing a good article properly eats most of a working day. There is the research, the structure, a full draft, then editing, fact-checking and formatting. Done right, a piece runs about six hours.
When the same person also does the client work and the business development, those hours are not there. The options narrow to writing less often, dropping the quality, or handing it to a writer who does not know the business.
What we built
A production flow where the operator drags a card across the board and the finished article turns up ready to read. The research, the drafting and the checks all happen first.
A failed check sends the draft back to revise, up to twice.
- 01
Move a card
The operator drags a card across the project board. That is the whole trigger.
- 02
It researches and drafts
The system researches the topic and writes a full draft in the brand voice, with an editing pass so it reads like a person.
- 03
Three independent checks
One looks at accuracy, one at search structure, one at whether it sounds human. A problem sends the draft back for a revision before anyone sees it.
- 04
A 30-minute review
The operator reads the finished draft, approves it or leaves a comment. About thirty minutes an article.
- 05
It publishes itself
Once approved, the article publishes on its own and the topic list updates so the angle is not repeated.
What we had to figure out
Early drafts were accurate but flat
The first version produced well-structured, correct articles that read like they were assembled rather than written. So we added an editing pass with one job: make it sound like a person, with shorter sentences and a real point of view. It runs before the checks.
One checker was bouncing almost everything
One review pass was failing nearly every article and sending it round again. We capped the revision cycles at two. If a draft still has issues after two passes, it goes to the operator with a note instead of looping forever.
The result
A good article ran about six hours from scratch, from research through to a final edit. That is now a 30-minute review, so each one saves 5 to 6 hours.
With the writing time gone, the only limit left is how long a review takes. Output went from 3 to 4 articles a month to 12 to 15.
One read of the finished draft and an approval. That is the whole job once the system has done the research, the draft and the checks.
The AI processing plus the automation platform, depending on how many articles run in a given month.
Figures from the system in production.
Talk through a problem like this one.
A free intro call. we'll show you where your hours are and what they're worth.