Automated Content Pipeline for Tech Marketing
RSS monitoring, relevance scoring, and AI draft generation for consistent multi-channel content.
Key Results
The Problem
Content marketing for a tech consultancy requires consistent output across multiple channels — LinkedIn, Twitter, blog — while staying relevant to fast-moving AI news.
Manual monitoring of 10+ sources, evaluating relevance, and drafting posts is time-consuming and inconsistent. Most teams either:
- Post inconsistently (gaps of days or weeks)
- Share irrelevant content that doesn't position the brand
- Spend hours on content that could be automated
The Solution
Automated content pipeline that handles collection, scoring, and draft generation while keeping humans in control of final output.
Collection — monitors 9 RSS sources including TechCrunch, Ars Technica, MIT Tech Review, Wired, The Verge, VentureBeat, Anthropic Blog, EU Digital Strategy, and Hacker News.
Scoring — each article is scored for business relevance using Claude API with company context. Returns relevance score (1-10), suggested channel, and angle for positioning.
Queuing — high-relevance items are queued with deduplication and expiry tracking.
Generation — channel-specific drafts (LinkedIn post, LinkedIn article, blog, Twitter, Reddit) following a detailed style guide with anti-patterns for AI-detectable writing.
Cross-posting — LinkedIn posts automatically adapt to Twitter teasers with --also-twitter flag.
Technical Highlights
Lockfile protection against hung cron jobs with 30-minute stale timeout.
Date-organized drafts — data/drafts/YYYY-MM-DD/ keeps outputs organized.
Style guide enforcement in prompts — no engagement questions, no "Here's the thing", no academic structure.
Fact-check markers — [VERIFY: ...] flags claims that need validation before publishing.
Telegram notifications on collection runs — know immediately when new content is ready.
Workflow
Automatic runs at 9:00 and 18:00 via cron: RSS feeds are fetched, articles scored, high-relevance items queued, and Telegram notification sent.
Manual workflow: review top articles, generate drafts for selected items, edit and publish.
Results
- 60-80 articles scored per day across all sources
- 10-15 high-relevance items queued for potential content
- 30 seconds draft generation time per post
- 2-5 minutes human editing time (minor tweaks to AI drafts)
What Makes It Work
Company context matters — the scoring prompt includes detailed company positioning, services, and voice guidelines. Better context = better relevance scoring.
Channel-specific output — LinkedIn, Twitter, and blog require different formats, lengths, and tones. One article can generate multiple appropriate drafts.
Human-in-the-loop — automation handles the grind, humans make the final call.
This system powers our own content marketing and is available as part of our Marketing Automation consulting.