Knowledge bases decay. Articles written six months ago reference deprecated tools, outdated processes, or people who have left the organization. Nobody is responsible for keeping everything current, so entropy wins. Employees learn to distrust the knowledge base and start asking questions in Slack instead.

The Problem

Most organizations have hundreds or thousands of knowledge base articles. No one person knows which are current and which are stale. Authors move to other teams or leave the company. The knowledge base becomes a liability rather than an asset because users cannot tell which information to trust.

The AI Approach

An LLM can audit knowledge base articles against current documentation, release notes, and organizational data to identify outdated content, suggest specific updates, and flag gaps where important topics have no coverage.

Three-Step Build

Step 1 - Content inventory. Export all knowledge base articles with metadata: title, author, last modified date, page views, and content. Identify articles not updated in the last six months as priority candidates for review.

Step 2 - Staleness analysis. For each candidate article, send it to an LLM along with recent changelog entries, product release notes, or process updates that might affect its accuracy. The model flags specific statements that may be outdated and suggests corrections.

Step 3 - Review routing. Generate a maintenance report listing articles needing updates, ranked by staleness severity and page views (high-traffic stale articles are the highest priority). Route each article to the appropriate team owner for review.

Where It Breaks

The model cannot verify facts against reality - it can only flag inconsistencies with other documents. Some articles are intentionally stable (company policies, legal terms) and should not be flagged as stale. Author metadata may be unreliable for routing.

The Production Path

Run the audit monthly. Track the percentage of flagged articles that actually needed updates to calibrate sensitivity. Add a content gap analysis that compares knowledge base topics against support ticket themes to identify missing documentation.