Content automation gets a bad rap because most people implement it wrong: they hit 'generate' and publish. The output is recognizably AI - generic, fluffy, and forgettable. Readers can tell. Google can tell. Your brand takes the hit.
The right version looks different. AI does the repetitive work - research aggregation, draft generation, formatting - while humans control strategy, editing, and final approval. The result is 5-10x more content with the same or better quality. Here's how we build it.
The content factory structure
Think of it as a pipeline with stages: briefing, generation, editing, formatting, and publishing. Automation handles briefing (pulling research and SEO data), generation (Claude writes the draft), formatting (applying brand templates), and publishing (posting to CMS via API). Humans handle strategy (which topics to cover), editing (review and refine the draft), and approval (nothing publishes without a human OK).
The briefing stage: what AI needs to write well
Poor AI content comes from poor briefs. A strong automated brief includes: target keyword and search intent, top 3 competing articles and what they cover, your unique angle (what you know that they don't), specific facts and examples to include, brand voice guidelines, and required length and structure. We build tools that auto-generate this from a topic input - pulling SEMrush or Ahrefs data, summarizing competitor content, and pre-filling your angle based on past content.
Getting Claude to write in your voice
The system prompt is everything. Before generating any content, we run a voice calibration process: feed Claude 5-10 of your best pieces and ask it to extract your writing patterns - sentence length, vocabulary, what you avoid, what you emphasize. These patterns become part of every generation prompt. The difference between 'write a blog post about X' and a 500-word system prompt with voice calibration is the difference between generic and publishable.
The human-in-the-loop editing step
We build the editing step directly into the workflow. Draft lands in Notion or Google Docs (writer's choice). It's flagged for review with an AI-generated edit checklist: 'Check these 5 specific things based on the brief.' The editor spends 20-30 minutes refining instead of 2-3 hours writing. When approved, one click triggers the publishing automation.
What the content pipeline automates:
- SEO research and keyword clustering
- Brief generation from topic input
- First draft (blog posts, product descriptions, social posts)
- Formatting for each channel (different length/style for LinkedIn vs email vs blog)
- CMS publishing via API (WordPress, Webflow, Notion)
- Social scheduling (Buffer, native APIs)
- Internal linking suggestions
- Performance tracking (which pieces drove traffic/conversions)
What we built for Persij
Persij (a fashion brand) needed product descriptions for 200+ SKUs per season, plus ongoing social content. We built a pipeline: CSV with product attributes → Claude generates description, bullet points, and 3 social variations → human reviews in batch (10 seconds per SKU) → approved content publishes to Shopify and schedules on social. What used to take a week of copywriting now takes 2 days of review.
Cost and build time
A basic content pipeline (topic → draft → review → publish) takes 5-7 days and costs $2,000-3,500. A full Content Factory (our productized version) includes multi-channel distribution, SEO optimization, and performance tracking - $4,500. This is the kind of system that replaces $3,000-6,000/month in freelance or agency costs.
We run the Content Factory as a productized service at 2pizza.team - you send voice notes or bullet points, we publish 30+ pieces a month in your voice. Or we build you the system. Both options available.
Free 30-min audit. We tell you what to automate first and what it would cost.