Turning Documentation Chaos Into a System: An AI-Driven Audit of a 200+ Product Content Library

Written by

in

, ,

A direct-to-consumer e-commerce client runs 170+ live sales funnels across several platforms — ClickFunnels, Shopify, a couple of internal tools — built up over years by multiple contributors. As the catalog grew past 200 product folders, the documentation that was supposed to keep it navigable hadn’t kept pace: fewer than 6% of product folders had any structured notes on what they were, what platform they lived on, or where to find their related tickets. A shared piece of code that every live order form depended on had three different internal documents disagreeing about which version was actually current — a silent risk to live sales pages.

What I Did

Using Claude Code as an AI research agent, I ran a full read-only survey of the entire content library — every product folder, every reference doc, every naming convention in use. Rather than proposing a heavyweight rebuild, I designed a lightweight standard: a small, required “context card” for every product that points to the client’s existing project-management system instead of duplicating its data, so there’s never a second copy of information to fall out of sync. I resolved the version conflict the same day it surfaced, then rolled the new standard out folder-by-folder, prioritizing whatever the team was actively working on rather than freezing operations for a big-bang migration.

The Result

  • Documentation coverage nearly quadrupled in the first rollout sprint, from under 6% to roughly 23% of the library, with backfilling ongoing
  • Resolved a live version conflict and cross-referenced the fix everywhere it was cited
  • Investigated two “duplicate ID” cases and correctly identified them as intentional legacy variants rather than errors, avoiding a rename that would have broken things

Why It Matters

Fast-growing catalogs, content libraries, and marketing systems all hit this same wall: the org chart of files and tools grows faster than anyone’s shared understanding of it. The fix isn’t more tooling — it’s a small amount of required structure that an AI agent can audit, maintain, and extend continuously, instead of a one-time cleanup that drifts again in six months.

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *