Content operations automation that keeps editors in charge

Content operations is everything between deciding to publish something and keeping it accurate once it is live: choosing topics, writing briefs, collecting expert input, drafting, editing, formatting, publishing and refreshing. Automation takes the clerical steps and leaves judgment with people.

This is how the stages split in the systems we build:

Every stage writes to one board, so anyone can see what is stuck and with whom.

A model handles the brief, the first draft and part of the check. Everything else is plain workflow automation, built with rules because rules are cheaper to run and easier to debug.

No, not for being AI-generated. Google's guidance on generative AI content (opens in a new tab), updated in October 2026, judges the result. It points to the scaled content abuse policy: using AI tools to "generate many pages without adding value for users" may violate it. The same page tells publishers to fact-check and review AI-generated content before it goes live.

That guidance shaped 4 rules in every pipeline we build.

1. Drafts start from something only you have. The expert input step runs before drafting, so each piece carries your numbers, examples and opinions. A pipeline fed only keywords produces exactly the commodity content Google's guidance describes.
2. A person approves every piece. The system creates drafts in your CMS and never publishes on its own. The approver's name is logged against each piece. This is what human in the loop means in practice.
3. Volume is set by review capacity. We size the pipeline to what your editors can check properly, and the board will not release more drafts than they have hours to review.
4. Numbers need sources. The checker flags any figure without a link, so the editor verifies it or cuts it.

AI-assisted pieces can rank when they meet the same bar as anything else: accurate, useful and specific to your company. If you also want those pages quoted in AI answers, the writing patterns are covered on our answer engine optimization page.

By writing the voice down as rules a machine can apply, then checking every draft against them. Most style guides are a page of adjectives such as confident, warm and expert. A model cannot act on adjectives. We turn your guide into 4 working parts:

• A rules file: sentence length range, person and tense, US or UK spelling, how numbers and prices are written and words you never use.
• 6 to 10 approved example pieces, chosen with your editor, that the drafting prompt reads as reference.
• A banned-phrase list checked by a script, so the same unwanted phrase is caught on its fortieth appearance as reliably as its first.
• A claims check that flags any number, comparison or customer reference without a source.

Then comes the feedback loop. When an editor changes a draft, the system stores the before and after. Once a month we review the most common edits and fold them into the rules file or the prompt. After a quarter, the edit log shows whether the same fixes still recur.

Voice also drifts when providers update their models. Before switching versions, we run a fixed set of 10 test briefs through the new model and compare the output with your approved examples.

One more rule: we do not let the model invent anecdotes or customer stories to make a draft feel personal. If a piece needs a real example, the expert input step asks for one.

Setup costs $8,000 to $25,000, fixed price. One content type, such as blog articles going from brief to WordPress drafts, sits near the bottom. Several content types, a headless CMS like Sanity or Contentful, expert interviews over Slack and a refresh queue sit near the top. If you want us to run it afterward, the operation plan costs $1,500 to $4,000 a month.

What moves the setup price:

• Content types. Each one, such as articles, case studies or product pages, needs its own brief template, examples and checks.
• The CMS. WordPress and Webflow drafts are quick to connect. Custom or headless setups take more field mapping.
• Approval steps. Legal or compliance review adds a stage, routing and an audit trail.
• Languages. Each language needs its own examples and its own reviewer.

Model usage from OpenAI or Anthropic is billed to your account at cost, with no markup, and the log shows it per piece.

On volume, the limit is review time. Drafting is fast; checking is not. As a planning example: if your editor has 10 hours a week for this and a piece takes 90 minutes to edit and approve, the pipeline can carry 6 or 7 pieces a week. More drafts than that only lengthen the queue. We measure your real edit time during the pilot and size the system from that number.

If you publish fewer than 4 pieces a month, a shared brief template and a checklist will do the job for a fraction of the price, and we will say so on the scoping call. The same goes if most of the effort is formatting and scheduling. That is a rule-based flow in Zapier or Make, which we build as a workflow integration for $1,500 to $6,000.

It is also the wrong choice when:

• Nobody inside the company can review for accuracy. The system makes reviewing faster. It cannot replace the reviewer.
• The goal is volume to fill a sitemap. Pages made in bulk without real input are what Google's scaled content policy targets. We decline that work.
• Your topics depend on original reporting or research. Interviews and data analysis are still human jobs. The board can track them, and adds little beyond that.

Not included in the setup price: writing your content as a standing service, running ads or social channels, and the search strategy that decides what to write. We do not sell general digital marketing.

If you need the strategy first, the SEO and GEO audit produces the question map this pipeline runs on. The AI search content engine blueprint shows the full system with each component and its build time.