All projects

AI-Powered Knowledge Base Automation

A multi-tool AI agent pipeline, orchestrated in n8n, that keeps a 1.47M-user knowledge base accurate — drafting, reviewing, and flagging content across its entire lifecycle.

AI Automation, n8n, Documentation

The problem

Documentation for a fast-moving product suite rots quietly. Every release, UI change, and new feature quietly invalidates something already published. With WPManageNinja’s Fluent suite shipping continuously to 1.47M+ users, keeping the knowledge base accurate by hand was a losing race — and the manual effort crowded out the work that actually mattered: good writing and good decisions.

What I built

I architected a multi-tool AI agent pipeline, orchestrated end-to-end in n8n, that treats documentation as a living system rather than a static archive.

  • Change monitoring — the workflow watches for product changes and surfaces what documentation they touch.
  • Drafting agents — AI agents (Claude, OpenAI Codex, Google AI Studio) generate updates and first-pass drafts in the house style.
  • Quality review — a review layer checks drafts for accuracy, clarity, and consistency before anything reaches a human.
  • Staleness detection — the system continuously flags outdated articles across the knowledge base so nothing silently goes wrong.
  • Human-in-the-loop publishing — everything lands as a review-ready draft; people make the final call.

The stack

n8n · Claude · OpenAI Codex · Google AI Studio · Markdown · custom prompt engineering and LLM integration

The outcome

The pipeline cut repetitive documentation effort dramatically, freeing the team to focus on high-value content decisions instead of chasing changelogs. It reframed the writer’s role from “keep everything current by hand” to “design the system that keeps everything current” — and earned AuthLab’s Noteworthy Contributor — Documentation recognition in 2025.