Case study 2026

Moon Beam Baker — Content Marketing Pipeline.

Automated content marketing system iterated through three versions, achieving ~75% cost reduction while producing multi-platform content.

Client
Moon Beam Baker
Sector
AI & Automation
Year
2026
Services
Multi-agent system design, LangChain orchestration, voice-aware content generation, research pipelines, multi-tenant architecture

Overview

Moon Beam Baker started as a content marketing automation project that became a proving ground for AI-powered content production. I designed and built an automated pipeline that handles the full content lifecycle — from news scanning and topic selection through research, writing, and publishing — iterated through three distinct architectural versions.

The Evolution

V1 — OpenClaw: Autonomous Agent Team

The first version used an autonomous multi-agent system where specialized AI agents handled different stages of the content pipeline. It worked — but the cost structure was unsustainable. Orchestration logic running through LLMs was the primary cost driver.

V2 — n8n: Deterministic Orchestration + LLM

The key insight: move orchestration to deterministic systems and reserve LLM calls for genuinely creative tasks. V2 rebuilt the pipeline in n8n with structured workflows, achieving a ~75% cost reduction while maintaining content quality. This version is live and reliable.

V3 — Code: Multi-Tenancy and Optimization

The current iteration moves the pipeline into code for full multi-tenancy support and further cost optimization. This allows the same infrastructure to serve multiple clients with isolated content strategies.

Pipeline Architecture

  1. News scanning — Automated monitoring of relevant industry sources
  2. Topic suggestion — AI-assisted topic selection based on content strategy
  3. Deep research — Comprehensive research gathering for each piece
  4. Writing brief — Structured brief generation for consistent output
  5. Voice-aware writing — Content produced in the brand’s specific voice
  6. AI-smell refinement — Post-processing to remove telltale AI patterns
  7. Human review — Quality gate before publishing
  8. Scheduled posting — Multi-platform distribution (LinkedIn, with Facebook and X planned)

Results

  • ~75% cost reduction from V1 to V2 by moving orchestration logic to deterministic systems
  • Multi-platform content production — consistent, on-brand output across channels
  • Scalable architecture — V3 supports multi-tenancy for additional clients