Content: 4 → 14 Posts/Week + LinkedIn + SEO Blog

The Starting Point: One Person, Four Posts, Zero Scale

When we looked at our own content operation, the numbers were sobering. One team member was responsible for everything: researching topics, writing copy, creating visual briefs, editing, and scheduling. The result was 4 Instagram posts per week — and even those felt rushed.

We had no LinkedIn presence at all. No SEO blog. Our potential clients couldn't find us through organic search, and we were invisible on the platform where B2B decisions actually happen. Content wasn't just a weakness — it was the ceiling on our growth.

Diagnosing the Bottleneck

We mapped the content creation process step by step. Each Instagram post took approximately 2.5 hours from idea to publication:

That's 10 hours per week for just 4 posts. Adding LinkedIn and a blog would have required hiring 1-2 more people — or finding a fundamentally different approach.

The Architecture: An AI Content Pipeline

We designed a multi-stage automation pipeline using n8n as the orchestration layer and Claude API as the content engine. Here's how it works:

Stage 1 — Topic Research. Every Monday, the system pulls trending topics from our industry, analyzes competitor content, and cross-references with our keyword strategy stored in Google Sheets. It generates 20+ topic suggestions ranked by relevance and SEO potential.

Stage 2 — Draft Generation. For each approved topic, Claude API generates a draft in our brand voice. We spent significant time engineering the prompts — feeding it our best-performing posts, brand guidelines, and audience personas. The system produces different formats: carousels for Instagram, long-form for LinkedIn, and SEO-optimized articles for the blog.

Stage 3 — Visual Briefs. For every post, the pipeline automatically generates a visual brief: color palette reference, text overlay suggestions, layout direction. This cut our designer's turnaround time in half.

Stage 4 — Review & Schedule. Drafts arrive in a Telegram bot for review. One tap to approve, or quick inline editing. Approved content is automatically scheduled across all platforms.

Week-by-Week Results

The Prompt Engineering That Made It Work

The biggest challenge wasn't the automation — it was making the AI output indistinguishable from human-written content. We iterated through 30+ prompt versions before landing on a system that consistently produced on-brand copy. Key principles:

• Feed the model 20 examples of our highest-performing posts as reference
• Include explicit brand voice guidelines: tone, banned words, preferred structures
• Use a two-pass system — first draft for content, second pass for tone and CTA optimization
• Different prompt templates for different platforms (Instagram carousel ≠ LinkedIn thought leadership)

What Changed Beyond the Numbers

The quantitative results speak for themselves: 3.5× content output, 3 channels instead of 1, 10+ hours per week freed up. But the qualitative shift was even more significant. Our content person went from being a "post factory" to a content strategist — reviewing, approving, and directing rather than writing every word from scratch.

Our SEO blog started ranking for target keywords within 6 weeks. LinkedIn became a genuine lead source, with inbound inquiries arriving weekly from posts that would never have existed without this pipeline.