Before we ever sold an editorial pipeline to a client, we built one for ourselves, because we had a problem that would not respond to hiring. A portfolio of publishing businesses, lead-generation sites in niches from international pet transport to close protection, alongside financial comparison platforms in the UAE, each needing research, articles, updates and technical upkeep. A conventional content operation at that breadth means an editorial floor. We do not have one.
The shape of the operation
Every site runs the same pipeline. Research workers watch the search landscape and the market and propose what should exist. Writing workers draft in each site's voice, grounded in checkable facts, with the claims sourced. Review workers check the drafts against written criteria: accuracy, style, structure, the rules about what never to say. What passes goes to publish. What fails goes back with reasons, and some of it stays rejected, which is the system working rather than failing.
Across the network this pipeline maintains a page count that has grown into six figures, built programmatically where the subject suits it and written piece by piece where it does not. Day to day it produces and updates content at a rate that would occupy a sizeable editorial team, and it does so on a schedule that does not depend on anyone's calendar.
The cost picture
The running cost has three parts, and all of them are small next to the salaries they replace. Model inference, which is metered per piece of work. Infrastructure, which for static publishing is modest. And review time, because a human editor still signs off where the criteria demand it, and that time is real.
We will not dress the comparison up with a precise multiple, because the honest version depends on what you count. The defensible statement is this: the fully loaded cost of a researched, reviewed, published article through this pipeline is a small fraction of the same article through a salaried team, and the gap widens as volume grows, because the pipeline's costs scale with usage while a team's costs scale with headcount.
Where the humans still matter
Direction is human. Which sites exist, which topics matter, what the standard is: those decisions do not belong to the pipeline. Judgement calls at the review gates are human wherever the blast radius justifies it. And the exceptions, the flagged anomalies and the genuinely new situations, land on a person by design.
The piece you are reading came through this pipeline, with a human editor signing off at the gate. We publish that fact deliberately. A firm selling autonomous publishing ought to be able to point at its own journal and say: this is the product, running on itself, in public.
From the Orwell Lab archive, rebuilt in our current design. The publication date above refers to the original article; market figures and provider details should be checked against current sources.
Discuss your own requirements ↗