AI glossary for businesses
Vector database
In short
A vector database is a database built to store embeddings, the lists of numbers that represent meaning, and to find the items closest to a query very quickly. Instead of matching exact words, it returns content that is similar in meaning. It is a common building block behind AI search, recommendations and RAG assistants.
How a vector database works
A normal database is like a filing cabinet sorted alphabetically. You find a customer by name or an order by number. A vector database is more like a librarian who shelves books by topic. Ask for "something on cutting delivery costs" and they walk to the right shelf, even if no title contains those exact words.
It works with embeddings. Each document, paragraph or product is turned into a list of numbers and stored. When a question comes in, it is also turned into numbers, and the database returns the nearest items. Special indexing makes this fast, even across millions of entries.
Well-known options include Pinecone, Weaviate, Qdrant and Chroma, as well as the pgvector extension for PostgreSQL. Many general platforms now add vector search as a feature.
In practice for a small business
Most SMEs never set up a vector database themselves. When you upload files to a ChatGPT project, a Claude project or Microsoft 365 Copilot, the tool handles this step for you.
You may need one when you build something custom. Example: a logistics company has 10,000 past support tickets and wants an assistant inside its own customer portal that suggests answers based on similar past cases. A developer would split the tickets into passages, create embeddings, store them in a vector database and connect it to a language model through RAG. At that point the choice of database, hosting region and update process becomes a real decision, so weigh carefully whether a custom build is worth it.
Why it matters and what to watch out for
Keep it in sync: if the source documents change but the database is not updated, the assistant gives outdated answers.
Access control: a vector database does not know who is allowed to see what unless you build that in. Plan permissions from the start.
Data location: check where the service stores data. EU hosting simplifies GDPR compliance.
Deletion requests: under GDPR, if a customer asks to be deleted, related entries in the vector database must go too.
Running cost: hosted services charge by storage and usage. For a few hundred documents, a built-in feature of an existing tool is often enough.