Who this is for
Businesses and agencies producing content on a schedule that keeps missing the schedule. Usually the writing is not the bottleneck; the bottleneck is gathering the research, formatting the brief, chasing approval, resizing the images and remembering to send the newsletter version. This service automates that surrounding work and uses a model for the first draft only. It sits under marketing and business automation services and uses the same guardrails as AI automation services.
It is a poor fit if nobody on the team can edit. A model produces confident, average prose that reads like every other model’s prose, and publishing that unedited will cost you more in credibility than it saves in hours. The workflow exists to give a competent editor a running start, not to remove the editor.
What gets automated
- Content briefs. A topic and target query go in; out comes a brief with the search intent, the questions to answer, suggested H2s, the internal links in and out, the word range and the call to action. The brief format is the one used across this site, so what is produced is directly usable.
- Research assembly. Existing pages on your own site, past posts on the topic, product details from your catalogue and the notes from a call transcript are collected into one document before drafting starts, so the draft is grounded in your material rather than the model’s guesswork.
- First drafts. Blog outlines, product descriptions, service page sections, email copy, ad variants and social captions, generated in your structure and your terminology from a style guide the workflow reads every time.
- Repurposing. One approved long piece becomes a LinkedIn version, an Instagram carousel outline, a newsletter section, a short script and a set of quote cards, each in the right length and tone rather than the same text pasted five times.
- Publishing pipeline. Approved content moves into WordPress as a draft with the title, meta description, category, tags, internal links and featured image slot already filled, then queues the social posts through social media automation.
- Housekeeping. Old posts flagged for refresh by age or traffic decline, broken internal links reported, and a content inventory kept current.
What I deliberately do not automate
- Publishing without an editor. Every workflow has a named approver and nothing reaches a live URL without them.
- Facts, numbers and claims. Statistics, prices, dates, client names and results are never model-generated. The workflow inserts them from your own records or leaves a visible gap for a person to fill.
- Bulk page generation. Producing hundreds of near-identical pages from a spreadsheet is how sites earn thin-content problems and cannibalise their own rankings. Where a templated set genuinely serves distinct searches, that is a deliberate content project, not a content automation.
- Anything under a byline that did not write it. If a person’s name is on the article, that person edits it.
- Comment and community writing. Replies that represent the business to a real customer are written by a human.
How a build runs
- Collect your best twenty pieces and the ones you rejected. These become the style reference and the test set. A workflow trained on your published voice produces drafts an editor can use; a workflow built on a generic prompt produces drafts an editor rewrites from scratch.
- Write the style and rules file: tone, spelling convention, banned filler phrases, terminology, what a claim needs before it can be written, what always requires a source.
- Design the pipeline on one page: trigger, research step, draft step, checks, approval, publish, distribution. You approve the diagram before I build.
- Build and measure. The draft workflow runs on ten past topics, and your editor scores how much rewriting each draft needed. That number decides whether the prompt is ready or needs another pass.
- Go live with a hard approval gate and a log of every input and output, so you can see what the model was given when a draft comes out wrong.
- Review monthly for drift after model updates, cost per piece and where the editor is still spending time.
What you receive
- Working workflows in your own n8n, Make or Zapier account with your own model API keys, so the cost and the data path are yours.
- The brief template, the style and rules file, and the prompts in a document you can edit without me.
- An approval queue in the tool your editors already use, with reject reasons captured for improving the prompts.
- Logging of every generation, a monthly API spend alert and a documented pause switch.
- Documentation and a handover session, plus optional monthly review of quality and cost.
Pricing pointer
Setup is quoted on request; a single pipeline, for example blog draft to WordPress with social repurposing, sits in the lower to middle part of the range, and the testing that makes the drafts usable is most of the work. Maintenance and prompt tuning are quoted on request, and model API usage is billed to your own account. Where the strategy itself needs work first, that is content marketing, and search-driven planning sits under content SEO.
Next step
Send me one recent piece you were happy with, one you were not, and the steps between deciding a topic and it being live. I will come back within four business hours with the pipeline diagram, what the model would and would not do in it, and a fixed price.
