Embedding is the process where a specialised model converts text (a sentence, paragraph, or document) into a vector – a list of numbers that captures the "meaning" of that text. Texts that are thematically similar have similar vectors.
Why does this matter? Embedding is the technology that enables semantic search – not searching for exact words, but for meaning. When a customer writes "my package didn't arrive" and your system contains a document about "complaints regarding undelivered shipments," embedding correctly connects them – even though they share no common words.
Embedding models are separate from generative LLMs. Popular ones: text-embedding-3-small by OpenAI, voyage-3 by Voyage AI, nomic-embed by Nomic.
Embeddings are stored in vector databases and are the foundation of every RAG system. This is the "invisible" technology that makes AI agents genuinely intelligent when working with your company documents.