Click to zoomOverview
A multi-path AI content pipeline built for marketing teams. Accepts a topic brief, classifies the content type, pulls live research via Tavily, drafts in structured chunks against a style rubric, and QA checks output before delivery. Handles long-form articles, short-form copy, and landing pages in a single workflow.
Stack
What needed solving
Content teams producing at scale struggle with consistency, research quality, and QA. Manual writing is slow, outsourced content is generic, and there is no system that checks its own output before delivery.
How I approached it
A self-checking AI content engine that classifies each brief, routes it through the right pipeline, researches live sources via Tavily, drafts section by section against a Google Docs style rubric, and runs a built-in QA pass before the content is delivered.
Step by step
- 01
Brief is received via chat message trigger
- 02
Classifier node determines content type: long-form article, landing page, or short-form copy
- 03
Switch node routes the brief to the correct pipeline
- 04
Tavily Search pulls live research relevant to the topic
- 05
Content is drafted in structured chunks against a style rubric stored in Google Docs
- 06
OpenAI and Claude run in parallel for drafting with a Merger combining outputs
- 07
A built-in Checker node QA reviews the draft before delivery
- 08
Final content is delivered via the same chat interface
Outcomes
- ✓
Long-form articles, landing pages, and short-form copy handled in one workflow
- ✓
Live research integrated into every piece via Tavily
- ✓
Built-in QA pass catches issues before content reaches the team
- ✓
Consistent output quality enforced by style rubric at every run
Key features
Dual LLM architecture with OpenAI and Claude running in parallel
52 nodes across multiple content paths in a single workflow
Self-checking system with built-in QA before delivery
Section-by-section drafting against a style rubric for consistent output