Train your knowledge

How the Content Library cleans and rewrites your content

In short: The Content library turns raw business information into structured, cleaned, deduplicated knowledge for agents. Updates apply automatically by default, with review tasks for changes that conflict with existing content or have low confidence.

A useful knowledge base does more than collect documents: it makes information consistent enough to use. Unless’s Living Knowledge pipeline processes material from business systems and turns it into a reference for agents. The steps below show how that transformation works.

From sources to a structured library

The process begins with raw information from business systems. Sources can include tickets, internal documents, Slack, Confluence, merge requests, recordings, and websites, so the incoming material may be scattered or repetitive.

Living Knowledge organizes, deduplicates, and cleans that material into the Content library. The Engine then continuously restructures and rewrites it into a non-ambiguous form, ready for agents to use.

How the library handles content

The library brings information together rather than leaving it in separate source systems. Signals flow into one structured library, where content is merged, deduplicated, and labeled.

This process is designed to address repeated information, contradictions, and different writing styles in raw sources. The result is a cleaner version of that material for agents to use, without surprises.

How changes are reviewed

Content library updates from Living Knowledge apply automatically by default. A review task appears when a change conflicts with existing content or has low confidence, so teams can review changes that need judgment.

A complete activity log records changes applied automatically. Teams can review recent changes and roll one back when needed, keeping the update process traceable.

Keeping knowledge current

The Content library is maintained as Living Knowledge processes new information, rather than relying on manual article edits or scripts. Teams can review suggested edits when they require attention, helping keep the library current and usable for agents.