AI and agents
Embeddings
An embedding is a list of numbers that a model produces to represent the meaning of a piece of text, an image or another item, so that similar items have similar numbers.
How it works in practice
An embedding model reads a passage and returns a vector, for example 1,536 numbers for OpenAI’s text-embedding-3-small. Passages about the same subject end up close together in that space even when they share no words. That is what lets software search, group or compare text by meaning.
Embeddings power semantic search, retrieval for RAG, duplicate detection and recommendations. Two practical points matter in a build. Every passage has to be re-embedded if you change embedding model, because vectors from different models cannot be compared. And embeddings capture meaning loosely, so searches for exact identifiers such as invoice numbers work better with ordinary keyword search alongside.
A worked example
A support team embeds every closed ticket. When a new ticket arrives, the system finds the five most similar past tickets and shows how each was resolved, which gives a new agent a head start on the answer.
Examples in this glossary describe typical builds for companies of 1 to 1,000 people. They are not named client results.















