Technology

Embeddings

A way of turning text into numbers that capture meaning, so computers can find documents that are about the same thing even when the words differ.

Embeddings convert text into long lists of numbers (vectors) arranged so that similar meanings land close together. 'Annual leave policy' and 'vacation rules' share almost no words, but their embeddings sit near each other — which is what makes semantic search possible: finding documents by meaning, not keyword.

Executives should care because embeddings are the invisible engine inside every good company knowledge assistant. When a RAG system 'searches your documents', it is embedding-based similarity doing the finding. This is why a well-built assistant answers questions phrased in ways your documents never used — and why keyword search always felt so dumb by comparison.

Concrete example: an employee asks the internal assistant 'how do I expense a client dinner?' The policy document says 'entertainment expenditure reimbursement'. Keyword search fails; embedding search finds it instantly, because the meanings match.

You will never manage embeddings directly. But knowing the concept lets you ask the right quality question of any AI knowledge project: how good is the retrieval, and how do we test it?