Embeddings

Working knowledgeCore AI Concepts

Also called: Vector Embeddings, Embedding

An embedding is a numerical representation of meaning, produced by converting a piece of text or an image into a list of numbers positioned so that similar meanings sit close together. Because the position is numerical, software can measure the distance between two of them and read that distance as similarity. This is what lets a search for "staff leaving" return a document about employee attrition that never uses the word "staff".

In practice

Embeddings are why an AI search tool finds the right document when nobody used the same words, and why it occasionally returns something confidently beside the point. When a vendor demonstrates search across your documents, ask what has been embedded and how often it is rebuilt, because a stale index answers from a version of the business that no longer exists.

Not sure where your organisation stands?

Take the free AI-readiness diagnostic.

Start the diagnostic